Background
Sporadic Alzheimer’s disease (AD) is the most frequent cause of dementia [
1] and is characterized pathologically by deposition of amyloid-beta (Aβ), hyperphosphorylated tau, and progressive neuronal dysfunction [
2,
3]. Pathological brain change in AD furthermore includes the increased accumulation of cerebral iron, which has been linked to several pathological processes associated with risk for AD and disease progression [
4‐
9].
Clinically, AD develops gradually and presents with progressive decline in multiple cognitive domains, particularly affecting episodic memory, executive functioning, and perceptual speed [
10‐
13]. While these alterations may also take place during normal aging, the concurrent incidence of subtle cognitive dysfunction and emerging AD brain pathology in the cognitively normal elderly is considered to reflect a preclinical stage of AD [
14,
15]. Moreover, progression of AD is significantly affected by genetic predisposition and presence of the apolipoprotein E (ApoE) ε4 allele, which is the strongest known genetic risk factor for late-onset AD [
16‐
20].
A variety of noninvasive neuroimaging techniques have been developed to assess AD pathology directly and AD-associated neurobiological changes. Positron emission tomography (PET) using the
11C-labeled tracer Pittsburgh Compound-B (PiB), is a validated method for measuring fibrillar Aβ [
21,
22] and has been used to assess pathological burden in several clinical studies [
23‐
25].
Cerebral iron can be measured in vivo by applying quantitative susceptibility mapping (QSM) magnetic resonance imaging (MRI) [
26,
27], which has very high signal-to-noise ratios (SNRs) when acquired at ultra-high field strength [
28,
29].
While the blood-oxygen-level dependent (BOLD) functional MRI (fMRI) contrast reflects in vivo neuronal activity [
30,
31], an ultra-high field strength of 7 Tesla (7T) may provide enhanced SNR and enriched contrast [
32‐
35]. Therefore, for the current study, a three-dimensional T2-weighted BOLD fMRI sequence was used (“T2-prep fMRI”), which was specifically designed for ultra-high magnetic field strength MRI acquisition [
36]. Functional connectivity, as inferred from synchronous fluctuations in activity in spatially distant brain areas [
37], is an established measure for investigating the integrity of functional brain networks and their potential impairment in AD [
38‐
41]. Investigation of “dynamic functional connectivity” additionally provides information on the expression of brain networks over time and has been used to characterize changes in brain network connectivity in neuropsychiatric disorders earlier [
42‐
45].
The primary aim of this study was to investigate whether dynamic expression of cognitive brain networks relates to interindividual variation of cognitive performance in healthy elderly subjects and their genetic predisposition for AD. Considering that “stationary” connectivity is significantly affected by local neurodegenerative brain change [
46‐
50], the second aim of this study was to examine the relationship between dynamic network expression and neuropathology.
11C-PiB-PET and QSM-MRI were thus used for measuring neuropathological burden, as indicated by the accumulation of cerebral Aβ and iron. QSM-MRI and resting state T2-prep fMRI were performed at an ultra-high field strength of 7T. Interindividual variability of cognitive performance over time was assessed by performing two neuropsychological tests for each participant, 2 years apart.
Discussion
By implementing dynamic functional connectivity analysis on ultra-high field strength MRI at 7T, reduced expression of a dynamic anterior-posterior brain network could be identified as a correlate of low episodic memory performance over time in cognitively normal elderly subjects. While strength of nodes implicated in this network related to mean regional susceptibility as a measure of local iron accumulation, no significant association could be observed for local Aβ plaque density, as inferred by PiB standardized uptake value ratio (SUVR). As dynamic functional connectivity changes relates to both lower episodic memory and ApoE-ε4, our findings may reflect brain change taking place in subjects at increased risk for AD and preclinical stages of AD, respectively. To our knowledge, this is the first study that demonstrates a relationship between memory decline within the normal range and altered dynamic network connectivity as a potential correlate of increased risk for AD in the healthy elderly.
Assessment of pathological burden included estimation of Aβ plaque density by administering
11C-PiB-PET, as has been demonstrated to be valid for characterizing progression of AD pathology previously [
21,
71‐
73]. Additionally, iron was measured by QSM-MRI [
26,
27] which was performed at ultra-high field strength for maximizing signal quality [
28]. While functional connectivity analysis of BOLD-fMRI data is a well-established measure of neural integrity in AD [
38,
39,
74], for the current study, T2-prep fMRI was used to avoid signal contortion near air cavities, but which nevertheless benefits from high SNR at 7T [
36]. Moreover, dynamic functional connectivity [
42] was assessed for inferring on the temporal expression of connectivity patterns by integrating information on both the regional extent and temporal evolution of coherent BOLD activity [
45,
64]. The studied population was cognitively assessed by performing tests and follow-up for language capacity, working memory, episodic memory, and executive function within 2 years. While neuropsychological assessment over time has been suggested previously to be a particularly reliable measure of cognitive performance in the elderly [
75,
76], the investigated study population in our study remained relatively stable regarding test performance within the study period. However, by splitting the study population by algebraic sign of yearly variability ratios, two subgroups that significantly differed regarding their rate of decline in the investigated cognitive domains could be identified that only showed moderate overlap regarding cognitive domains affected by lower performance over time. Some participants in our study performed better at follow-up, which may be explained by practice effects as reported previously for longitudinal studies on cognitively normal elderly subjects [
77].
Our finding of an association between memory performance and dynamic connectivity appears consistent with a concatenation of earlier reports on altered functional connectivity in AD [
50,
78,
79] as well as associations between distinct cognitive impairment and increased AD risk [
15,
80]. Central nodes of the dynamic anterior-posterior network found to be associated with episodic memory performance exhibited increased iron for the lower episodic memory group. The strongest effects were observable for the left precuneus, right caudate, and right anterior cingulate. This observation appears consistent with previous reports on subcortical regions being primarily affected by iron accumulation in neurodegenerative brain disorders [
81,
82]. While no differences in local Aβ plaque density were measured as being associated with network dynamics, a distinct impact of ApoE-ε4 on network expression was observable, suggesting an association with an increased risk for AD. Moreover, cerebral iron accumulation may reflect pathological processes implicated in AD [
81,
83‐
85], and local interactions of accumulated iron have been suggested previously to promote neuronal damage in the context of AD [
6,
7,
85‐
88]. These earlier considerations on interactions between iron and AD pathology may be consistent with our lack of identifying significant associations between local iron and network-dynamics, when iron was corrected for Aβ.
While our data indicate memory decline within the normal range as a potential correlate of altered dynamic network connectivity and increased genetic risk for AD, our findings might furthermore support earlier considerations on the relevance of pathological processes reflected by local iron, such as oxidative stress, free radical activity, and mitochondrial dysfunction [
83,
84]. These processes may be reflected by functional changes, primarily affecting brain regions with high metabolic activity and increased susceptibility to age-related damage [
89]. Considering these pathological processes as secondary to the earlier manifestation of AD pathology, they nevertheless may substantially contribute to cognitive decline [
4,
81,
90,
91] and may thus represent a correlate of the well-established phenomenon of functional disconnection in AD [
39,
50,
89]. This interpretation may be consistent with previous considerations on a stronger association of functional impairment with secondary pathological processes than with Aβ plaque density itself [
92,
93].
The following limitations have to be allowed for when appraising our reported findings. While neuropsychological performance was assessed based on measures within 2 years, neuroimaging was performed only once and thus only confers cross-sectional information. Additional longitudinal studies are necessary to investigate the temporal relationship of the different constituents of pathological burden, which included Aβ plaque density and iron load in the current study. As the number of study participants affected by cognitive decline, and thus power to identify functional correlates of low performance, varied between the domains investigated, negative findings for language capacity, working memory, and executive function need to be interpreted with caution. Moreover, while MRI at ultra-high field strength may provide advantages in SNR and thus facilitate detection of pathological change [
28], reproducibility of findings may be difficult as it requires the implementation of sequences that were originally performed on 7T on more readily available clinical scanners with lower field strength.
Conclusions
While the association between memory decline in the elderly and emerging AD-related pathology is well established, our findings suggest that variation in the subclinical range of memory performance may be linked to alterations in functional network dynamics. Moreover, our data suggest that altered network dynamics reflect regional pathological burden, as characterized by increased iron accumulation, and also genetic risk, as conferred by ApoE-ε4. Additional studies are necessary to clarify whether the observed dynamic functional changes reflect impaired neural integrity and thus possibly a symptom-free stage of incipient cognitive disorder, or alternatively may represent adaptive mechanisms activated for maintenance of brain functionality during aging.
Acknowledgments
We thank all subjects for their participation in our study. We thank Linjing Mu, Ph.D. and Geoff Warnock, Ph.D., from the Division of Nuclear Medicine, University of Zürich, Switzerland, for their help in generation of 11C-labeled Pittsburgh Compound-B tracer for positron emission tomography (Linjing Mu) and calculation of the cortical PiB ratio scores (Geoff Warnock).