Background
Alzheimer’s disease (AD) with concomitant cerebrovascular disease (CeVD) is a leading cause of age-related cognitive impairment [
1]. Such a mixed pathology is not only associated with distinct neurodegenerative patterns, but also with greater cognitive decline and earlier dementia onset than AD or CeVD only [
2‐
4].
The network-based degeneration hypothesis suggests that the disease-related spread of degeneration follows a pattern based on existing brain networks [
5‐
8]. Emerging evidence illustrates that AD and mild cognitive impairment (MCI) are associated with functional connectivity (FC) and structural connectivity (SC) alterations in the default mode network (DMN) with associated memory impairment, while CeVD shows FC and SC changes in the executive control network (ECN) [
9‐
13]. Recent findings from our group using single DMN/ECN seeds indicate differential neural network changes that may be reflective of different underlying pathology in subjects with and without CeVD [
14]. However, most studies have used single seed-based approaches to assess FC changes in concomitant CeVD and AD. Thus, given the multiple DMN and ECN core regions and accumulative evidence on seed-dependent FC patterns, such a region-based effect of CeVD on their network connectivity in AD and amnestic MCI (aMCI) using simultaneous FC and SC approaches remains to be elucidated [
6,
7,
15,
16]. Furthermore, increased vascular burden could influence cognition through network dysfunction via impaired SC [
2,
17‐
19]. Indeed, CeVD markers have been associated with cognition in MCI [
20‐
22]. However, the effect of CeVD on functional and structural network connectivity in AD needs further investigation, especially in aMCI [
22,
23].
Given these gaps, we aimed to concurrently assess FC and SC changes within and between the DMN and ECN in aMCI and AD subjects with and without CeVD and their associations with cognitive decline using a multiple seed-based approach. We hypothesized that non-CeVD groups would show DMN FC damage underlying memory impairment while CeVD participants would show ECN FC damage underlying attention and executive function impairment. Such network divergence patterns would be less evident in SC; instead, SC disruptions are likely to be more severe in CeVD than non-CeVD.
Discussion
A hypothesis-driven multiple seed-based approach and combination of functional and structural connectivity analyses were used to assess the effect of CeVD on DMN and ECN connectivity in aMCI and AD patients. We demonstrated region-specific FC changes in AD patients with and without CeVD, which related to cognitive impairment. Both AD and AD+CeVD subjects showed reductions in hippocampal FC within the DMN. However, parietal and medial prefrontal-parietal DMN FC was increased in CeVD groups but decreased in AD subjects. As predicted, intra-ECN alterations in frontal and frontoparietal FC were observed most extensively in CeVD subjects. Notably, aMCI+CeVD subjects exhibited similar intra-network FC changes to AD+CeVD, while aMCI subjects did not show any intra-network FC changes compared with HCs. Inter-network FC reductions were observed in AD and AD+CeVD subjects, while aMCI and aMCI+CeVD subjects primarily showed increases when compared with controls. Direct comparisons between CeVD and non-CeVD groups revealed disease severity-dependent alterations in inter-network FC with decreased DMN-ECN FC in aMCI+CeVD compared with aMCI but increased DMN-ECN FC in AD+CeVD compared with AD. Moreover, intra-DMN FC changes were associated with cognitive impairment primarily in non-CeVD groups while ECN-related FC changes were associated with cognitive impairment primarily in CeVD groups. Additionally, CeVD groups had greater SC damage within and between the two networks compared with non-CeVD groups at both aMCI and AD stages. Similar to our FC findings, aMCI with CeVD but not those without CeVD had SC declines. This study suggests that subjects with CeVD show distinct network FC phenotypes and severe SC deterioration in the brain which underlie cognitive impairment.
The DMN is important for cognitive functions such as episodic memory and has been widely implicated in AD [
5,
7]. Our non-CeVD and CeVD AD patients showed extensive intra-DMN FC alterations. However, posterior DMN FC alterations involving the posterior cingulate, precuneus, and hippocampus seeds were dominant in AD subjects, as observed previously [
12,
61]. These regions have been shown to comprise the core DMN as well as being involved in early amyloid deposition and associations with autobiographical and episodic memory [
5,
62]. In support of such findings, intra-DMN FC and cognition associations were primarily observed in non-CeVD groups in our study. Additionally, AD subjects showed increases in frontal FC which were negatively associated with cognition, thus indicating that such increases were derogatory in nature [
63,
64]. On the other hand, intra-DMN medial prefrontal-parietal FC was decreased in AD subjects but increased in both aMCI and AD with CeVD. Such a divergence in FC changes between CeVD and non-CeVD subjects could possibly be due to disruption of frontal pathways in the presence of vascular disease [
65]. Indeed, associations between intra-ECN frontal FC increase and frontal SC decrease were found in AD groups in our study (Additional file
1: Supplementary Results section 2.6). Thus, while AD subjects both with and without CeVD showed similar involvement of hippocampal FC, medial prefrontal-parietal FC was instead differentially targeted in CeVD and non-CeVD, likely indicative of differential subnetwork FC alterations in the presence of CeVD.
Widespread intra-ECN FC alterations including increases in frontal FC were observed in AD+CeVD subjects, possibly reflecting greater influences on ECN connectivity in CeVD [
11,
14]. We also found associations between higher frontal ECN FC and higher WMH volume in both aMCI and AD groups (with and without CeVD) and postulate that such increases in ECN FC could be representative of CeVD abnormalities in the brain (Additional file
1: Figure S5). Additionally, such increases in frontal FC were associated with worse executive, attention, and memory function in subjects with CeVD, indicating a derogatory influence. Parietal ECN FC was reduced in CeVD subjects and was associated with worse attention function. Indeed, associations between markers of CeVD (WMH and lacunes) and executive/attention function have been demonstrated [
2,
11,
18]. Moreover, task-based fMRI studies in the healthy elderly with CeVD and resting-state fMRI studies in vascular cognitive impairment have shown alterations in ECN connectivity [
66]. Importantly, associations between ECN FC and cognition were primarily observed in subjects with CeVD in our study. Thus, in line with previous studies, our findings further lend evidence to the influence of concomitant AD and CeVD on network FC and cognition.
Furthermore, findings from our group and others show inter-network segregation as being consistently affected in AD patients and point towards its role in cognition [
49,
67]. Interestingly, we observed lower DMN-ECN frontoparietal FC in aMCI+CeVD compared with aMCI, but higher frontoparietal FC in AD+CeVD compared with AD subjects. Such differential inter-network FC changes at the aMCI and AD stages likely provide some evidence for stage-dependent alterations in network segregation in the presence of CeVD. While reductions in aMCI+CeVD inter-network FC possibly reflect a compensatory mechanism in the presence of CeVD, increased inter-network FC with disease progression to AD+CeVD might reflect a breakdown in inter-network segregation possibly due to CeVD-related neuronal loss and degradation of white matter networks [
68].
Prior FC studies have demonstrated inconsistent findings regarding disruptions in MCI [
12,
21]. For example, whole-brain FC studies have shown both FC decreases and increases in parietal and temporal regions, reflecting a concurrent state of impairment as well as compensation [
61,
63]. In this study, intra-network FC alterations were observed in aMCI+CeVD subjects when compared with controls, which largely mirrored alterations observed in AD+CeVD subjects [
14]. Interestingly, no intra-network FC alterations occurred in the aMCI-only subjects. This indicates that aMCI+CeVD subjects appear to be further along the disease spectrum than non-CeVD aMCI subjects. We speculate that the absence of FC changes might also reflect a possible compensatory mechanism accompanied by network reorganization in aMCI, which may breakdown in the presence of CeVD [
61,
63]. Further studies integrating task-based and task-free FC methods are required to study how CeVD influences whole-brain network topology and its relationship with cognitive impairment in aMCI.
In concordance with our FC patterns of large-scale alterations, our findings indicated that, overall, CeVD groups showed more widespread SC changes compared with non-CeVD groups [
17]. Importantly, SC disruptions in CeVD groups occurred primarily along intra-ECN (i.e., frontal or fronto-parietal connections such as between the DLPFC and PPC). Our observations are supported by prior studies showing decreased frontal and parietal nodal efficiency in CeVD and its mediating effect on frontal lobe structure and cognition [
13]. Direct comparisons between aMCI and AD subjects with and without CeVD also highlight greater intra-ECN SC damage. Crucially, we found early intra-ECN and inter-network SC damage with sparing of intra-DMN fibers in aMCI+CeVD subjects, in agreement with prior studies [
13,
17]. As observed in our seed-based FC analyses, these differences in SC were not observed in the non-CeVD aMCI subjects. Such findings indicate an ECN-specific structural and additive influence of CeVD that likely begins in aMCI. Additionally, and unlike FC, there was no dissociation in the SC-cognition relationship between CeVD and non-CeVD groups. Performance on both memory and executive/attention domains was associated with intra-network SC in both AD and aMCI subjects with and without CeVD, indicating that white matter damage might lead to deficiencies in both memory and executive/attention domains regardless of CeVD status [
20,
21]. Our SC findings reflect that CeVD may be associated with greater white matter degeneration and lend evidence to the additive hypothesis regarding the influence of CeVD when there is concomitant AD [
3,
19].
Our study has some limitations. As a hypothesis-driven seed-based approach was chosen to compare SC and FC in the two networks of interest, these findings may be affected by inter-subject anatomical variability. A relatively large proportion of the CeVD subjects in our study had infarcts in the frontal regions (Additional file
1: Table S8), which may bias the ECN functional connectivity estimation and associations with cognition. Additionally, although groups were not age-matched and disease duration was not available, age differences were accounted for in all analyses and disease severity was matched between CeVD and non-CeVD groups at the aMCI and AD stages, respectively. It has also been suggested that probabilistic fiber tracking can be influenced by the presence of WMH in the brain [
69]. While we did control for WMH volume in our structural connectivity statistical analyses, WMH may still confound the fiber tracking results, especially in the crossing-fiber regions. Furthermore, out of the 248 subjects included in our study, only 45 (9 HC, 16 aMCI, 14 aMCI+CeVD, and 6 AD) had amyloid imaging data. Thus, we are unable to assess how the heterogeneity in the etiology of the patient groups, especially at the aMCI stage, would have influenced our findings.