A comprehensive assessment of resting state networks: bidirectional modification of functional integrity in cerebro-cerebellar networks in dementia

Abstract

In resting state fMRI (rs-fMRI), only functional connectivity (FC) reductions in the default mode network (DMN) are normally reported as a biomarker for Alzheimer's disease (AD). In this investigation we have developed a comprehensive strategy to characterize the FC changes occurring in multiple networks and applied it in a pilot study of subjects with AD and Mild Cognitive Impairment (MCI), compared to healthy controls (HC). Resting state networks (RSNs) were studied in 14 AD (70 ± 6 years), 12 MCI (74 ± 6 years), and 16 HC (69 ± 5 years). RSN alterations were present in almost all the 15 recognized RSNs; overall, 474 voxels presented a reduced FC in MCI and 1244 in AD while 1627 voxels showed an increased FC in MCI and 1711 in AD. The RSNs were then ranked according to the magnitude and extension of FC changes (gFC), putting in evidence 6 RSNs with prominent changes: DMN, frontal cortical network (FCN), lateral visual network (LVN), basal ganglia network (BGN), cerebellar network (CBLN), and the anterior insula network (AIN). Nodes, or hubs, showing alterations common to more than one RSN were mostly localized within the prefrontal cortex and the mesial-temporal cortex. The cerebellum showed a unique behavior where voxels of decreased gFC were only found in AD while a significant gFC increase was only found in MCI. The gFC alterations showed strong correlations (p < 0.001) with psychological scores, in particular Mini-Mental State Examination (MMSE) and attention/memory tasks. In conclusion, this analysis revealed that the DMN was affected by remarkable FC increases, that FC alterations extended over several RSNs, that derangement of functional relationships between multiple areas occurred already in the early stages of dementia. These results warrant future work to verify whether these represent compensatory mechanisms that exploit a pre-existing neural reserve through plasticity, which evolve in a state of lack of connectivity between different networks with the worsening of the pathology.

Keywords: Alzheimer disease; cerebro-cerebellar networks; functional connectivity alterations; mild cognitive impairment; resting state fMRI.

Figure 1

The 15 RSNs identified in the present investigation. Auditory network (AN), right and left ventral attention networks (VAN), language network (LN), sensory-motor network (SMN), executive control network (ECN), default mode network (DMN), medial visual network (MVN), lateral visual network (LVN), cerebellar network (CBLN), salience network (SN), task positive network (TPN), basal ganglia network (BGN), frontal cortical network (FCN), anterior insula network (AIN). In this and the following figures, the 3D scans are projected onto three representative sections chosen on the sagittal (left), coronal (middle) and axial (right) planes.

Figure 2

Areas with reduced FC in the RSNs. Images show areas with reduced FC (p ≤ 0.05, TFCE-corrected) in AD compared to HC (light blue), in MCI compared to HC (blue), and in both AD and MCI compared to HC (dark blue). Note that the areas of reduced FC are mainly located in the prefrontal (PreF) and precentral (PreC) areas. Moreover ADs present a larger number of areas with reduced FC than MCI suggesting a more extended functional corruption in AD than MCI.

Figure 3

Areas with increased FC in the RSNs. Images show areas with increased FC (p ≤ 0.05, TFCE-corrected) in AD compared to HC (yellow), in MCI compared to HC (orange), and in both AD and MCI compared to HC (red). Note that the cerebellum presents large areas of increased FC in MCI, suggesting active compensatory mechanisms.

Figure 4

Ranking of gFC changes in MCI and AD. In order to better identify the patterns of gFC changes within the networks, for each contrast (MCI < HC, MCI > HC on the top; AD < HC, AD > HC on the bottom of the picture) we ranked the RSN alterations in terms of their decreasing gFC. We identified 6 different prototypical patterns: (1) FCN (green arrows) showed the largest gFC reduction, both in AD and MCI, and no increase in any conditions; (2) AIN (turquoise arrows) showed no gFC reduction, either in AD or MCI, but it showed a large increase in both conditions; (3) DMN (blue arrows) showed the second largest gFC reduction, both in AD and MCI, but at the same time it showed also the largest increase in both conditions; (4) BGN (orange arrows) shows a reduction in AD but not in MCI, while it also shows an increase in both cases; (5) CBLN (magenta arrows) shows just a minor decrease in AD but a remarkable increase in MCI only; (6) LVN (yellow arrow) shows gFC increase in both AD and MCI, with stronger effect in AD. At the same time it showed a small gFC reduction in MCI, but not in AD.

Figure 5

Patterns of rgFC changes within specific RSNs. Pictures show the value of rgFC across conditions (MCI < HC and MCI > HC on the left side, AD < HC and AD > HC on the right one) for each sub-area of the brain: cerebellum (C), prefrontal (PreF), limbic (L), occipital (O), parietal (P) precentral (PreC) and temporal (T) for the networks that showed specific prototypical patterns: DMN, FCN, AIN, BGN, CBLN, and LVN.

Figure 6

Global cerebellar alterations in AD and MCI. For each RSNs and for each contrast (AD < HC, AD > HC, MCI < HC, and MCI > HC), the histograms show the global changes occurring in the cerebellum C(rgFC) in AD (A) and in MCI (B). For each RSN, the set of histograms shows: (1) the global changes of the rgFC index in the cerebellum (C), (2) the contribution to the rgFC in the cerebellum given by the spatial extension of the clusters of functional alteration (N_tstatFC/N_RSN), (3) the contribution of the magnitude extension of the cerebellar functional alteration [meanFC(C)_Pwith p = MCI or AD, normalized for the to the meanFC(C)_HC].

Figure 7

Comparison of structural and functional changes in AD and MCI. (A,B) VBM analysis. AD subjects compared to HC (A) show large areas of GM atrophy (orange) in bilateral mesial-temporal network. MCI subjects compared to HC (B) show sparse areas of GM atrophy (orange) in left and right hippocampus, left amygdala, parahippocampal gyrus, precuneus and superior temporal gyrus. AD show more extended areas of atrophy than MCI. (C,D) FC analysis. Both AD and MCI compared to HC show large clusters of FC reductions (blue) mainly localized in the prefrontal areas involving VAN, DMN, DAN, FCN, SN, and SMN, with more extended areas of FC reductions in AD than MCI. (E,F) VBM/FC matching. Matching VBM and FC maps of FC reduction for AD compared to HC and MCI compared to HC. The overlapping between atrophy and FC changes (red clusters) is very limited.

Figure 8

Correlation between rgFC changes and clinical and neuropsychological tests. Results are reported for each contrast (A) AD < HC, (B) MCI < HC, (C) AD > HC and (D) MCI > HC. Correlation matrixes (left) and the corresponding statistical significance of correlations (right) between clinical and neuropsychological tests (1/TMT-A, 1/TMT-B, MMSE, MP, CF, SF, ROCF-copy, and ROCF-rec) and rgFC areas: cerebellum (C), prefrontal (PreF), limbic (L), occipital (O), parietal (P), precentral (PreC), and temporal (T). Correlation and anti-correlation are reported on different color scales. Correlations that satisfied a statistical threshold of p < 0.01 are marked with a star (^*). The statistical significance (matrices on the right) of each Pearson's correlation coefficient (matrices on the left, with the positive correlations in red scale and the negative ones on blue scale) is represented as 1–p-value.