Decoupling of Local Metabolic Activity and Functional Connectivity Links to Amyloid in Alzheimer’s Disease

Abstract

Background:

Both ongoing local metabolic activity (LMA) and corresponding functional connectivity (FC) with remote brain regions are progressively impaired in Alzheimer’s disease (AD), particularly in the posterior default mode network (pDMN); however, it is unknown how these impairments interact. It is well known that decreasing mean synaptic activity of a region, i.e., decreasing LMA, reduces the region’s sensitivity to afferent input from other regions, i.e., FC.

Objective:

We hypothesized progressive decoupling between LMA and FC in AD, which is linked to amyloid-β pathology (Aβ).

Methods:

Healthy adults (n=20) and Aβ+patients without memory impairment (n=9), early MCI (n=21), late MCI (n=18) and AD (n=22) were assessed by resting-state fMRI, FDG-PET, and AV-45-PET to measure FC, LMA, and Aβ of the pDMN. Coupling between LMA and FC (r_LA/FC) was estimated by voxelwise correlation.

Results:

R_LMA/FC decreased with disease severity (F=20.09, p<0.001). This decrease was specifically associated with pDMN Aβ (r=−0.273, p=0.029) but not global Aβ (r=−0.112, p=0.378) and with the impact of Aβ on FC (i.e., r_Aβ/FC, r=−0.339; p=0.006). In multiple regression models r_LMA/FC was also associated with memory impairment, reduced cognitive speed and flexibility, outperforming global Aβ, pDMN Aβ, pDMN LMA, and pDMN FC, respectively.

Conclusion:

Results demonstrate increasing decoupling of LMA from its FC in AD. Data suggest that decoupling is driven by local Aβ and contributes to memory decline.

Keywords

Alzheimer’s disease amyloid-β default mode network functional connectivity local metabolic activity

INTRODUCTION

In Alzheimer’s disease (AD), both local metabolic activity (LMA) and functional connectivity (FC) between remote brain regions are progressively impaired in several brain regions. For example, ¹⁸fluor-desoxyglucose positron emission tomography (FDG-PET) reveals progressive hypometabolism [1 –4] primarily in posterior regions of the default mode network (pDMN, i.e., medial and lateral parietal areas [5, 6]. Cortical signaling requirements (i.e., mainly action potentials and postsynaptic effects) dominate cortical energy demand. More specifically, ca. 80% of energy consumption in the cortex is dedicated to synaptic signaling, while the remaining energy is consumed by housekeeping functions [7, 8]. Therefore FDG-PET hypometabolism estimates loss of mean local synaptic activity in AD. FC (i.e., statistically significant co-activity across time) of ongoing blood oxygenation level dependent (BOLD) fluctuations of resting-state functional MRI (rs-fMRI) is progressively reduced in the pDMN along the spectrum of AD [9 –11]. Resting-state BOLD fluctuations reflect fluctuations in local neural activity at slow frequency [12], therefore reductions in BOLD-FC estimates are believed to estimate loss of FC among LMAs of remote regions. While reductions in both LMA and FC spatially overlap within the pDMN, it is poorly understood how both impaired processes interact in AD. The current study aims at studying the relationship of LMA and FC in AD, how it relates to amyloid-β (Aβ), and whether it is relevant for cognitive decline.

LMA and FC are closely related. It has been demonstrated that LMA signals of FDG-PET are highly correlated with FC signals of rs-fMRI across voxels, indicating the tight physiological coupling between LMA and FC. Furthermore, such voxel wise correlation of LMA and FC (r_LMA/FC) seems to be sensitive to the cognitive state of the subject [13]. Based on this evidence, we stated as our first hypothesis that in patients with AD, r_LMA/FC is reduced and that such reduction is associated with cognitive decline.

One of the pathological hallmarks of AD is increasing Aβ pathology [14]. Studies in transgenic mice demonstrate on the one hand, that Aβ interacts with neuronal activity, resulting in significant portions of hyperactive neurons [15]. Complementary, across patients, regional mean LMA (i.e., measured by FDG-PET) correlates positively with local Aβ plaque load, suggesting that mean LMA increases with local Aβ accumulation [16]. On the other hand, animal studies show that Aβ impairs FC across the cortex via desynchronizing slowly fluctuating LMA of remote regions [17]. Therefore, due to its impact on both LMA and FC, we stated in our second hypothesis that local Aβ is relevant for suggested decoupling of LMA from FC.

To test these hypotheses, healthy adults and Aβ positive (Aβ+) patients, ranging from cognitive normal state to mild cognitive impairment (MCI) and dementia, who were assessed by rs-fMRI, FDG-PET, and AV-45-PET to measure FC, LMA, and Aβ, respectively, were identified from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. The pDMN was defined by rs-fMRI of an independent sample of healthy controls. R_LMA/FC was estimated by spatial voxelwise correlation between LMA and FC across the pDMN. R_LMA/FC was associated with both Aβ levels and cognitive scores (i.e., delayed recall performance).

Conceptualized as a proof of principle study and based on previous findings showing the pDMN to be the primary network affected by hypometabolism, Aβ deposition, and FC reduction [6], we focused our analysis of de-coupled LMA and FC in AD on the pDMN. Other resting-state networks, which are also affected by FC reduction [11], hypometabolism and Aβ load [6] as well as complementary questions such as whether the de-coupling between LMA and FC is a function of network are not topic of the current study.

Table 1

Group comparisons for age, sex, APOE4 status and cognitive performance and mean DMN scores of FC, FDG-SUVR and Florbetapir SUVR, and whole brain amyloid SUVR

	CN Aβ- n = 20	CN Aβ+ n = 9	EMCI n = 21	LMCI n = 18	AD n = 22	ANOVA
						F	p
Age (y)	74.6	76.3	71.2	73.6	73.3	1.01	0.41
(SD)	(7.8)	(5.6)	(5.6)	(7.3)	(7.9)
Sex Female/Male	14/6	5/4	9/12	8/10	12/10		0.44^†
(Female % /Male %)	(70/30)	(55.6/44.4)	(42.9/57.1)	(55.6/44.4)	(54.5/45.5)
APOE4 allele(carriers/noncarrieres)	4/15	1/7	(14/5)	(6/8)	(17/2)		0.000^†
(carriers % /noncarriers %)	(21.1/79.9)	(12.5/87.5)	(73.7/26.3)	(42.9/57.1)	(89.5/10.5)
Education (y)	16.1	16.1	15.4	16.5	15.1	0.84	0.50
(SD)	(2.2)	(1.4)	(2.7)	(2.4)	(2.7)
ADAS cog global	5.90	6.00	9.10	13.33	24.41	39.14	0.000
(SD)	(2.89)	(1.80)	(4.15)	(5.65)	(8.29)
Delayed recall subscore	14.10	14.00	8.29	3.22	1.36	114.81	0.000
(SD)	(2.94)	(1.23)	(2.03)	(2.71)	(1.76)
Trail Making Test A	34.7	31.1	31.2	38.1	69.9	12.70	0.000
(SD)	(10.1)	(8.0)	(7.8)	(14.0)	(33.8)
Trail Making Test B	70.0	74.8	101.6	110.8	195.1	12.09	0.000
(SD)	(44.9)	(20.7)	(40.9)	(71.6)	(69.3)
pDMN FC z-scores	1.84	1.80	1.85	1.76	1.74	3.42	0.012
(SD)	(0.11)	(0.12)	(0.11)	(0.11)	(0.12)
pDMN FDG-PET SUVR	1.65	1.49	1.57	1.46	1.21	19.51	0.000
(SD)	(0.17)	(0.14)	(0.17)	(0.20)	(0.17)
pDMN AV45-PET SUVR	1.03	1.34	1.32	1.37	1.46	32.51	0.000
(SD)	(0.04)	(0.13)	(0.15)	(0.15)	(0.14)
Whole brain AV45-PET SUVR	1.02	1.36	1.37	1.41	1.54	12.42	0.000
(SD)	(0.05)	(0.16)	(0.17)	(0.19)	(0.43)

^†χ² p; CN Aβ-, cognitively normal amyloid negative; CN Aβ+, cognitively normal amyloid positive; EMCI, early mild cognitive impairment; LMCI, late mild cognitive impairment; AD, Alzheimer’s disease; ANOVA, analysis of variance; ADAS cog, Alzheimer’s Disease Assessment Scale– cognitive; SD, standard deviation; pDMN, posterior default mode network; FC, functional connectivity; FDG-PET, ¹⁸fluor-desoxyglucose positron tomography; SUVR, standard uptake value ratio; AV45-PET, florbetapir positron emission tomography.

MATERIALS AND METHODS

Data source

Data used in the preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu). The ADNI was launched in 2003 as a public private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD [18 –20].

Participants and imaging data

All subjects who finished cognitive assessment, rs-fMRI, a T1-weighted structural MRI, Florbetapir PET and FDG-PET within 3 months were identified. The final sample consisted of 90 participants: 20 Aβ- cognitively normal healthy controls (CN Aβ-), 9 Aβ+cognitively normal healthy controls (CN Aβ+), 21 patients with early MCI (EMCI) Aβ+, 18 patients with late MCI (LMCI) Aβ+and 22 patients with AD Aβ+(demographics in Table 1). Levels of MCI (early or late) are determined using logical memory testing results (http://adni.loni.usc.edu/study-design/background-rationale/). Furthermore, to control for potential confounding effects of APOE status on coupling between local activity and FC, we extracted the APOE status from the ADNI-database when available and dichotomized it into APOE4-carriers (at least one APOE4 allele) and non-carriers (no APOE4 allele), following the approach of Sheline and co-authors [21].

More details are found in the Supplementary Material and on the ADNI website (http://adni.loni.usc.edu/methods/).

Cognitive assessment

Cognitive performance was assessed by the Alzheimer’s Disease Assessment Scale-cognitive (ADAS-cog) [22], with scores range from 0 to 70. As proxy for memory performance, we focused on the Delayed Word Recall subscores. Furthermore, we used Trail Making Tests Part A and B to test for motor and visual search speed as well as set-shifting and inhibition as proxies for non-memory cognitive functions [23].

Preprocessing and independent component analysis

Statistical Parametric Mapping Version 8 (SPM8, http://www.fil.ion.ucl.ac.uk/spm/) and Data Processing Assistant for Resting State fMRI (DPARSF, http://rfmri.org/DPARSF) were used for data preprocessing [24]. After discarding the first three functional images of every subject, we used rigid coregistration to motion-correct the resting state fMRI data. To control for motion-induced artifacts, point-to-point head motion and mean head motion were estimated for each subject [25]. Excessive head motion (cumulative translation or rotation >3<mm or 3° and mean point-to-point translation or rotation >0.15<mm or 0.1°) was applied as an exclusion criterion. Framewise displacement [25] was not different across groups (p<> <0.05, ANOVA and post-hoc t-tests; Supplementary Table 1). The structural image was coregistered to the mean functional MRI image (using rigid registration), segmented into gray matter, white matter and cerebrospinal fluid, and registered to the Montreal Neurological Institute (MNI) stereotactic space template using unified segmentation. After nuisance regression of head motion, white matter, cerebrospinal fluid, and global signal, the same normalization parameters were then used to normalize the functional images into MNI space with an isotropic voxel size of 3x3x3 mm³. Normalized functional images were finally smoothed with an 8x8x8 mm³ Gaussian kernel.

To control for the possible contribution of a change in temporal signal to noise ratio (tSNR) across the AD spectrum, temporal SNR of fMRI data was assessed. For each subject, the voxelwise mean time series was divided by the voxelwise standard deviation of the time series. The resulting tSNR images were masked by the pDMN template and entered into a one-way ANOVA to assess whether a main effect of tSNR across the five groups could be found. There was no significant difference between the groups in terms of tSNR (F=5.09, p=0.75). This result suggests that our findings about group changes in resting state fMRI-derived measures including coupling outcome r_LMA/FC are not affected by SNR changes across groups.

Preprocessed data were decomposed into 35 spatially independent components reflecting intrinsic brain networks via a group-level independent component analysis (ICA) framework [26], using the Group ICA of fMRI Toolbox (GIFT, http://www.icatb.sourceforge.net). To automatically select the pDMN, we applied multiple spatial regression analyses on the 35 independent components using publicly available network templates from a multicenter study [27]. Then, for each subject an individual independent component reflecting the pDMN was derived via back-reconstruction. The component was composed of the network’s time course and spatial map of z-scores. Voxels with high FC within the network have high z-scores, whereas voxels that are not part of the network have z-scores near 0. Individual levels of pDMN FC were derived from voxels with z-values ≥1 and with a gray matter probability ≥0.5. A z-threshold of 1 was set in order to omit negative z-values reflecting voxels negatively correlated to pDMN activity but to ensure that the whole extent of the pDMN was included in the analyses (see Fig. 1A).

Fig. 1

Default mode network and group differences in mean functional connectivity (FC), amyloid-β (Aβ) plaques, and local metabolic activity (LMA). A) Mask of the posterior default mode network used in the current study. B) Group differences across AD stages in mean FC, Aβ pathology, and LMA. Patients had progressive loss of LMA (as measured by FDG-SUVR, ANOVA, F=19.51, p <0.001, post hoc tests) and FC (as measured by z mean score ANOVA, F=3.42 p=0.012, post hoc tests). Patients had progressive increase of Aβ (as measured by AV-45-SUVR, ANOVA, F=32.51, p <0.001, post hoc tests). CN Aβ-, cognitively normal amyloid negative; CN Aβ+, cognitively normal amyloid positive; EMCI, early mild cognitive impairment; LMCI, late mild cognitive impairment; AD, Alzheimer’s disease; SUVR, standardized uptake value ratio; ⁺p <0.10, *p <0.01, **p <0.001.

To ensure that our analysis is restricted to grey matter of pDMN, we created a gray matter mask, which was then combined with the pDMN mask. More specifically, gray matter segmented images derived from anatomical data were used to create a subject-specific binarized gray matter mask with a probability threshold of 0.5.

Individual masks (i.e., the multiplication of pDMN-mask and individual grey matter mask) were included in all further analyses.

PET data preprocessing

Both subjects’ FDG-PET and AV-45-PET scans were rigidly coregistered to their corresponding structural MRI image and then normalized on the MNI template using the previously acquired normalization parameters. This procedure resulted in structural MRI, resting-state fMRI and PET data with identical dimensions and in the same space. For each subject and FDG-PET, all voxel values were converted to SUVRs by scaling to the vermis of cerebellum as reference [1]. Additionally, images were smoothed using an 8x8x8 mm³ Gaussian kernel. For each subject and AV-45-PET scans, all voxel values were converted to SUVRs by scaling to the cerebellum as reference. Images were smoothed using an 8x8x8 mm³ Gaussian kernel. In order to derive individual measures of Aβ load, individual AV-45-SUVRs were averaged across those voxels belonging to the individual pDMN mask (i.e., voxels with z-value ≥1, and gray matter probability ≥0.5).

Outcome measures

To determine the coupling between LMA and FC for each subject, we calculated the voxel-wise or spatial correlation r_LMA/FC between FDG-SUVR and z-scores across the pDMN. To be independent of the impact of Aβ on coupling between LMA and FC, we chose a partial correlation approach using voxel-wise AV-45-SUVR as additional variable. To estimate the impact of Aβ on both LMA and FC, we calculated corresponding r_Aβ/LMA and r_Aβ/FC scores within the same voxel-wise partial correlation frame work. To control for potential impact of grey matter atrophy, we repeated our analysis using voxel-wise grey matter signal of T1-weighted images as nuisance variable.

Statistical analysis

To compare variables across groups, separate ANOVAs were computed for FDG-SUVR, FC z-scores, AV-45-SUVR, and r_LMA/FC, respectively. To estimate the association between r_LMA/FC and memory decline in patients (i.e., Aβ+subjects; CN Aβ+, EMCI Aβ+, LMCI Aβ+, AD Aβ+), we performed regression analyses with delayed recall scores of ADAS-cog as dependent variable and r_LMA/FC, pDMN’s Aβ, LMA, and FC as independent variables. Additional nuisance variables were age, sex, and years of education. The same analyses were also performed for non-memory cognitive scores, namely Trail Making Tests Part A and B, to control whether the association between r_LMA/FC and memory decline is specific for memory function.

To estimate the relevance of pDMN’s Aβ on the coupling between LMA and FC in patients, r_LMA/FC was correlated with both pDMN’s mean AV-45-SUVR and global AV-45-SUVR, respectively, by the use of partial correlation with age, sex, years of education, and APOE-status as nuisance variables. Correlation with global AV-45-SUVR was performed in order to test the specificity of potential results. The same approach was used to estimate the relevance of the impact of Aβ on LMA and FC, respectively, measured by r_Aβ/LMA and r_Aβ/FC for the coupling between LMA and FC, r_LMA/FC.

RESULTS

LMA, FC, and Aβ in the pDMN

Sample characteristics for age, sex, years of education, ADASCog, and Trail Making Test scores are depicted in Table 1. Groups did not differ in age, sex and years of education. We found significant group differences in the APOE-status. Patients (i.e., Aβ+subjects; CN Aβ+, EMCI Aβ+, LMCI Aβ+, AD Aβ+) had progressively impaired delayed recall. The number of voxels of the pDMN did not differ across groups (Supplementary Table 1). Patients showed a significant group effect in LMA as measured by FDG-SUVR (Table 1 and Fig. 1B, ANOVA, F=19.51, p <0.001). Patients had increased Aβ as measured by AV-45-SUVR in the pDMN (Table 1 and Fig. 1B, ANOVA, F=32.51, p <0.001) and in the whole brain (Table 1, ANOVA, F=12.42 p <0.001). Patients also showed a small but significant group effect in FC as measured by z mean score of the pDMN (Table 1 and Fig. 1B, ANOVA, F=3.42 p=0.012).

Progressive decoupling between LMA and FC is associated with memory decline

All subjects had positive coupling r_LMA/FC (Fig. 2A, Table 2). Across groups, r_LMA/FC was progressively decreased (Fig. 2A, Table 2, ANOVA, F=20.09, p <0.001). This result persisted when voxelwise grey matter intensity values were included as nuisance variable (Table 2, ANOVA, F=19.83, p <0.001).

Fig. 2

Coupling between local metabolic activity (LMA) and its functional connectivity (FC) across AD stages and its link with memory decline. A) The voxel-by-voxel spatial correlation between LMA and FC (i.e., r_LMA/FC) declines along the AD spectrum, starting in LMCI and progressing in AD (ANOVA, F=20.09, p <0.001, post hoc tests). B) Delayed recall subscores of the ADASCog (Alzheimer’s Disease Assessment Scale-cognitive; Rosen et al. [22]) is significantly and specifically associated with r_LMA/FC (univariate regression analysis, β=<0.568, p <0.001). CN Aβ-, cognitively normal amyloid negative; CN Aβ+, cognitively normal amyloid positive; EMCI, early mild cognitive impairment; LMCI, late mild cognitive impairment; AD, Alzheimer’s disease; *p <0.01, **p <0.001.

Table 2

Results of within subject multimodal correlation analyses across groups. R values are means

	Whole sample n = 90	CN Aβ- n = 20	CN Aβ+ n = 9	EMCI n = 21	LMCI n = 18	AD n = 22	ANOVA
							F	p
r_LMA/FC	0.18	0.27	0.27	0.25	0.12	0.05	20.09	0.000
(SD)	(0.13)	(0.08)	(0.12)	(0.08)	(0.11)	(0.10)
r_LMA/FC (GM^*₎	0.18	0.27	0.27	0.25	0.12	0.06	19.83	0.000
(SD)	(0.13)	(0.08)	(0.12)	(0.08)	(0.11)	(0.10)
r_Aβ/FC	0.02	-0.04	– 0.02	0.02	0.05	0.07	5.51	0.000
(SD)	(0.10)	(0.06)	(0.07)	(0.11)	(0.08)	(0.09)
r_Aβ/FC (GM^*₎	0.02	-0.05	-0.03	0.01	0.04	0.07	5.40	0.000
(SD)	(0.09)	(0.06)	(0.07)	(0.11)	(0.07)	(0.09)
r_Aβ/LMA	0.32	0.20	0.37	0.35	0.33	0.32	4.35	0.003
(SD)	(0.16)	(0.10)	(0.18)	(0.15)	(0.19)	(0.16)
r_Aβ/LMA (GM^*₎	0.32	0.20	0.37	0.35	0.33	0.32	4.43	0.003
(SD)	(0.16)	(0.10)	(0.18)	(0.15)	(0.19)	(0.16)

CN Aβ-, cognitively normal amyloid negative; CN Aβ+, cognitively normal amyloid positive; EMCI, early mild cognitive impairment; LMCI, late mild cognitive impairment; AD, Alzheimer’s disease; GM, grey matter; SD, standard deviation; FC, functional connectivity, LMA, local metabolic activity, Aβ, amyloid-β; r values are partial correlation coefficients for FC, Aβ, and LMA (and GM) across voxels of the posterior default mode network; ^*covariate of no interest.

Univariate regression analyses revealed that in patients, delayed recall performance was significantly associated with r_LMA/FC (β=<0.538, p <0.001, Fig. 2B) but also with global Aβ (β=<−0.333, p=0.014), pDMN Aβ (β=<− 0.346, p=0.011) and pDMN LMA (β=<0.502, p <0.001); there was a strong trend in the correlation with pDMN FC (β=<0.268, p=0.066). To test for specificity, multiple regression analysis was performed with delayed recall score as dependent variable and r_LMA/FC, LMA, FC, global Aβ, and pDMN Aβ as independent variables. This analysis revealed a significant association between recall performance and r_LMA/FC only (β=<0.541, p=0.008), while all other independent variables did not reach significance: global Aβ (β=<− 0.207, p=0.097), pDMN Aβ (β=<− 0.101, p=0.560), pDMN LMA (β=<0.303, p=0.166), pDMN FC (β=<− 0.229, p=0.178). All regression models included age, sex and years of education as additional nuisance variables. In the same way, univariate regression analyses also revealed a significant association between r_LMA/FC and Trail Making Tests A (β=<− 0.396, p=0.002) and B (β=<− 0.418, p=0.002), respectively. In a multiple regression model including global Aβ, pDMN Aβ, pDMN LMA, pDMN FC as additional independent variables, r_LMA/FC was not associated with results in Trail Making Tests A (β=<− 0.169, p=0.342) and B (β=<− 0.229, p=0.223), respectively.

Decoupling between LMA and FC is related with local Aβ

Next we studied how Aβ interferes with decoupling between LMA and FC in patients. First we tested whether particularly local Aβ (i.e., Aβ in the pDMN) is associated with r_LMA/FC in patients. Partial correlation analysis, controlling for age, sex, years of education, and APOE-status, revealed a significant negative correlation between r_LMA/FC and DMN Aβ (Fig. 3A, r=− 0.239, p=0.044) but not for global Aβ (Fig. 3B, r=− 0.170, p=0.122). This result suggests that Aβ within the pDMN is relevant for the decoupling of LMA and FC within the pDMN.

Fig. 3

Decoupling of functional connectivity (FC) from local metabolic activity (LMA) and its relation with amyloid-β (Aβ). A) Significant correspondence between coupling of FC with LMA (r_LMA/FC) and Aβ pathology load of the posterior default mode network (pDMN; p=0.029; partial correlation analysis controlling for age and sex), but not with the (B) global brain’s Aβ (p=0.378; partial correlation analysis controlling for age and sex).

Then we tested whether the link of Aβ with LMA and FC, respectively, is associated with the decoupling between LMA and FC. By the use of a partial correlation approach, we found significant correlation between r_Aβ/FC and r_LMA/FC (r<₌<− 0.329; p=0.021) but not between r_Aβ/LMA and r_LMA/FC (r=0.100, p=0.493). This result suggests that the link between Aβ and FC within the pDMN is relevant for the decoupling of LMA and FC within the pDMN, but not the link between Aβ and LMA. In other words, the suggested impact of Aβ on FC within the pDMN is important for the decoupling of FC from LMA but not the impact of Aβ on LMA.

DISCUSSION

By the use of multimodal imaging and cognitive assessment in Aβ positive patients along different stages of AD, we tested the hypothesis of a weakening of physiological coupling between LMA measured through glucose metabolism and FC in AD. LMA was progressively decoupled from FC in patients, and this decrease of correspondence was associated with memory decline. To the best of our knowledge, our result provides first evidence for impaired physiological coupling of LMA and FC in AD. Furthermore, reduced correspondence was specifically linked with Aβ in the pDMN but not with Aβ of the whole brain, suggesting a potential role of local Aβ in decoupling LMA from FC. In addition, the voxel-by-voxel link of Aβ and FC but not with LMA was associated with decoupling LMA from FC, suggesting a potentially desynchronizing effect of Aβ on LMA-FC correspondence.

The coupling of LMA and corresponding FC is a basic feature of brain physiology. Specifically, linking microcircuit activity and blood oxygen level dependent signals via biophysical models, macroscopic FC appears to rest upon inter-regional coupling of averaged ensemble activity and particularly of low-level spiking activity [28, 29]. Recent imaging studies support this model; for example, our group demonstrated recently that, at rest, LMA measured by FDG-PET determines inter-regional FC to a high degree (20− 50%) expressed via regional spatial correlations between LMA and FC [13]. The current study translates this finding to AD patients. We found that the correspondence between FC and LMA is progressively reduced in Aβ positive patients across the clinical severity spectrum of AD. In line with several previous studies, we found—for input data for our coupling measure—both progressive decrease of LMA [6] and moderate decrease of FC within the pDMN in AD [6, 11]. Progressive decoupling of LMA and FC cannot be explained by both distinct pDMN size and atrophy across groups, as we controlled for these two variables. Our finding is supported by previous findings, across patients, of correlations between hypometabolism and loss of (global) FC in areas of the pDMN [30]. One should note that our approach is based on within subject spatial correlations between LMA and FC, which estimate subject’s physiological coupling of LMA with FC.

Furthermore, the loss of correspondence between LMA and FC in the pDMN was associated with cognitive decline in terms of delayed recall, cognitive speed, set shifting, and inhibition. The association between verbal memory and coupling between LMA and FC was specific, as multiple regression analysis identified coupling as the only variable associated with recall deficits while other variables such as Aβ or LMA alone were not related. The pDMN is well known to be involved in cognitive processes, particularly in recall processes via interactions between posterior parietal lobes and medial temporal lobes [31]. As the correspondence between LMA and FC outperforms other variables of pDMN impairment such as LMA or FC alone, it might be of high interest whether the coupling between LMA and FC is especially relevant for such recall performance. To a lesser extent, reductions in coupling of pDMN between LMA and FC were also unspecifically associated with cognitive speed, set shifting, and inhibition. These domains are not attributed to the functions of the pDMN. We suggest that the degree of coupling of pDMN between LMA and FC is foremost associated with memory performance and unspecifically associated with general cognitive decline.

Decoupling of LMA and FC was associated with the Aβ plaque level of the pDMN. This finding was specific for Aβ within the pDMN, since we did not find any correlation with global Aβ levels. This result suggests that local Aβ is relevant for decoupling of LMA and functionality. This is supported by studies in transgenic mice, which show that both Aβ plaques and peptides influence LMA of neuronal ensembles and its synchrony with remote activity [15, 17]. Imaging studies in AD patients also demonstrate an aberrant effect of local Aβ on FC by reducing FC anywhere in an affected network as a function of local Aβ plaque concentration [32, 33]. We furthermore showed that the local voxel-by-voxel correspondence between Aβ and FC was specifically associated with decoupling of LMA and FC. This association was specific for the voxel-by-voxel correspondence between Aβ and FC, because voxel-by-voxel correspondence between Aβ and LMA was not significantly related to coupling between LMA and FC. The local correspondence between Aβ and FC or LMA, respectively, was estimated via spatial correlations of Aβ with LMA or FC across the pDMN. This approach reflects how variability of Aβ across voxels co-varies with variability of FC or LMA, respectively. Although the voxel wise interactions are hard to interpret in a simple way, we suggest both measures to reflect—as supported by animal experiments investigating the impact of Aβ on neuronal processes—the impact of Aβ on LMA and FC, respectively. Independently from interpretation, the specific association between Aβ and FC and its significant correlation with coupling of LMA and FC suggests that the relation between Aβ and FC, but not the relation between Aβ and LMA, is relevant for the decoupling of LMA from FC. Regarding the correspondence between Aβ and LMA, we found robust positive scores across patients, which are increased in patients. This means increasing Aβ is linked with increasing LMA, which is in line with both previous findings in patients with MCI [16] but also with animal studies of Aβ-induced hyperactivity (always having in mind some caution by such translational interpretations) [15]. Therefore, Aβ-induced increase of LMA is not relevant for decoupling of LMA and FC. On the other hand, the interaction between Aβ and FC is associated with decoupling of LMA and FC. As animal studies demonstrate that Aβ desynchronizes ongoing LMA of remote brain regions, a simple explanation of our finding would be that, despite of its relative hyperactive local effect, Aβ desynchronizes FC between remote brain regions via these local activity changes. As animal studies indicate further that both hyperactive and desynchronizing effects of Aβ are mediated by impairing local GABA-ergic inter-neurons [17], we speculate further that the effects of Aβ on GABA-ergic inter-neurons might contribute to basic decoupling of LMA from FC. It is clear that these speculations have to be further tested by studies in both animals and patients.

Our results may indicate an interesting observation of potentially increased coupling between LMA and FC in patients with MCI. One should be clear that the mild increase of coupling between LMA and FC does not reach statistical significance. But two types of previous findings in humans support such potential coupling increase in early AD: on the one hand, several studies report increased FC in early stages of the disease (e.g., [34, 35]). On the other hand, some studies report increased local activity, also in early disease stages [16, 36]. Due to the suggested coupling model [37] that increased local activity induces increased FC, one might expect that increases in local activity in patients might be accompanied by increases in FC, with both being reflected by an increased coupling between local activity and FC. One might further speculate that local activity increases reflect increased proportions of spontaneously hyperactive neurons induced by Aβ peptides, as shown in animal models of Aβ [15]. Altman and colleagues demonstrated for example for patients with MCI and AD, that increasing levels of Aβ deposits are associated with increasing local activity [16]. Future studies may focus on the specific link between increased local activity, corresponding FC and their coupling in AD.

Some limitations need to be considered. First of all, our study focuses on the pDMN, since it is the primary network affected by hypometabolism, Aβ deposition [6], and FC reduction in AD [11]. However, pathological changes in AD follow a gradient across intrinsic FC networks [11, 32]. Across AD stages, LMA-decrease and Aβ deposition show a pronounced pattern in the DMN followed by frontoparietal networks, which motivates the extension of this study on other higher-cognitive networks.

Second, the large number of Aβ positive patients increases the power of our findings. The distribution of the correlation coefficient (LMA and FC) in subgroups of patients is of remarkable variability. Several non-exclusive cases might explain this variability: the small sample sizes of the subgroups, both LMA and FC measures are of high intra- and interindividual variability, so the correlation of both might just reflect this variability. LMA and FC were not assessed simultaneously and were acquired at distinct spatial resolutions, potentially accounting for the high level of variability in inter-modality correspondence. Third, our approach does not account for further AD-relevant pathologies such as tau pathology, which might also influence the physiological coupling between LMA and FC within regions of the pDMN. Fourth findings in AD suggest a non-linear trajectory of intrinsic FC across pre-symptomatic and symptomatic stages. FC was found to be increased in patients at risk [34] and in very early MCI [36], while decreases were consistently demonstrated at later stages in patients with late MCI [35] and manifest AD [10]. In our sample, we found an increase in both FC and LMA from a presymptomatic (cognitively normal Aβ positive) to an early symptomatic stage (early MCI). Although of general interest, the effects did not reach significance, possibly because of the small sample size.

In summary, our findings indicate increasing decoupling of LMA from FC in AD, which is associated with memory decline and local Aβ.

Footnotes

ACKNOWLEDGMENTS

Christian Sorg is funded by the DFG (SO 1336/1-1), German Federal Ministry of Education and Science (BMBF BMBF 01ER0803) and the Kommission für Klinische Forschung, Technische Universität München (KKF 8765162).

Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer's Association; Alzheimer's Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer's Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Chételat

, Desgranges

, Landeau

, Mézenge

, Poline

, de la Sayette

, Viader

, Eustache

, Baron

(2008) Direct voxel-based comparison between grey matter hypometabolism and atrophy in Alzheimer’s disease. Brain 131, 60–71.

La Joie

, Perrotin

, Barré

, Hommet

, Mézenge

, Ibazizene

, Camus

, Abbas

, Landeau

, Guilloteau

, de La Sayette

, Eustache

, Desgranges

, Chételat

(2012) Region-specific hierarchy between atrophy, hypometabolism, and β-amyloid (Aβ) load in Alzheimer’s disease dementia. J Neurosci 32, 16265–16273.

Mevel

, Desgranges

, Baron

, Landeau

, De la Sayette

, Viader

, Eustache

, Chételat

(2007) Detecting hippocampal hypometabolism in mild cognitive impairment using automatic voxel-based approaches. Neuroimage 37, 18–25.

Mosconi

(2013) Glucose metabolism in normal aging and Alzheimer’s disease: Methodological and physiological considerations for PET studies. Clin Transl Imaging 1, 4.

Raichle

(2015) The brain’s default mode network. [Review]. Annu Rev Neurosci 38, 433–447.

Grothe

, Teipel

, Alzheimer’s Disease Neuroimaging Initiative (2016) Spatial patterns of atrophy, hypometabolism, and amyloid deposition in Alzheimer’s disease correspond to dissociable functional brain networks. Hum Brain Mapp 37, 35–53.

Attwell

, Laughlin

(2001) An energy budget for signaling in the grey matter of the brain. J Cereb Blood Flow Metab 21, 1133–1145.

Attwell

, Iadecola

(2002) The neural basis of functional brain imaging signals. Trends Neurosci 25, 621–625.

Greicius

, Srivastava

, Reiss

, Menon

(2004) Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: evidence from functional MRI. Proc Natl Acad Sci U S A 101, 4637–4642.

10.

Sorg

, Riedl

, Mühlau

, Calhoun

, Eichele

, Läer

, Drzezga

, Förstl

, Kurz

, Zimmer

, Wohlschläger

(2007) Selective changes of resting-state networks in individuals at risk for Alzheimer’s disease. Proc Natl Acad Sci U S A 104, 18760–18765.

11.

Nuttall

, Pasquini

, Scherr

, Sorg

(2016) Degradation in intrinsic connectivity networks across the Alzheimer’s disease spectrum. Alzheimers Dement (Amst) 5, 35–42.

12.

, Snyder

, Zempel

, Smyth

, Raichle

(2008) Electrophysiological correlates of the brain’s intrinsic large-scale functional architecture. Proc Natl Acad Sci U S A 105, 16039–16044.

13.

Riedl

, Utz

, Castrillón

, Grimmer

, Rauschecker

, Ploner

, Friston

, Drzezga

, Sorg

(2016) Metabolic connectivity mapping reveals effective connectivity in the resting human brain. Proc Natl Acad Sci U S A 113, 428–433.

14.

Braak

, Braak

(1991) Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 82, 239–259.

15.

Busche

, Eichhoff

, Adelsberger

, Abramowski

, Wiederhold

, Haass

, Staufenbiel

, Konnerth

, Garaschuk

(2008) Clusters of hyperactive neurons near amyloid plaques in a mouse model of Alzheimer’s disease. Science 321, 1686–1689.

16.

Altmann

, Ng

, Landau

, Jagust

, Greicius

, Alzheimer’s Disease Neuroimaging Initiative (2015) Regional brain hypometabolism is unrelated to regional amyloid plaque burden. Brain 138, 3734–3746.

17.

Busche

, Kekuš

, Adelsberger

, Noda

, Förstl

, Nelken

, Konnerth

(2015) Rescue of long-range circuit dysfunction in Alzheimer’s disease models. Nat Neurosci 18, 1623–1630.

18.

Jagust

, Bandy

, Chen

, Foster

, Landau

, Mathis

, Price

, Reiman

, Skovronsky

, Koeppe

, Alzheimer’s Disease Neuroimaging Initiative (2010) The Alzheimer’s Disease Neuroimaging Initiative positron emission tomography core. Alzheimers Dement 6, 221–229.

19.

Langbaum

, Chen

, Lee

, Reschke

, Bandy

, Fleisher

, Alexander

, Foster

, Weiner

, Koeppe

, Jagust

, Reiman

, Alzheimer’s Disease Neuroimaging Initiative (2009) Categorical and correlational analyses of baseline fluorodeoxyglucose positron emission tomography images from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Neuroimage 45, 1107–1116.

20.

Mueller

, Weiner

, Thal

, Petersen

, Jack

, Jagust

, Trojanowski

, Toga

, Beckett

(2005) Ways toward an early diagnosis in Alzheimer’s disease: the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Alzheimers Dement 1, 55–66.

21.

Sheline

, Morris

, Snyder

, Price

, Yan

, D’Angelo

, Liu

, Dixit

, Benzinger

, Fagan

, Goate

, Mintun

(2010) APOE4 allele disrupts resting state fMRI connectivity in the absence of amyloid plaques or decreased CSF Aβ42. J Neurosci 30, 17035–17040.

22.

Rosen

, Mohs

, Davis

(1984) A new rating scale for Alzheimer’s disease. Am J Psychiatry 141, 1356–1364.

23.

Reitan

, Wolfson

(1985) The Halstead-Reitan neuropsychological test battery, 2 ed. Neuropsychology Press, Tucson, AZ.

24.

Chao-Gan

, Yu-Feng

(2010) DPARSF: A MATLAB Toolbox for “Pipeline” Data Analysis of Resting-State fMRI. Front Syst Neurosci 4, 13.

25.

Van Dijk

, Sabuncu

, Buckner

(2012) The influence of head motion on intrinsic functional connectivity MRI. Neuroimage 59, 431–438.

26.

Calhoun

, Adali

, Pearlson

, Pekar

(2001) Spatial and temporal independent component analysis of functional MRI data containing a pair of task-related waveforms. Hum Brain Mapp 13, 43–53.

27.

Allen

, Erhardt

, Damaraju

, Gruner

, Segall

, Silva

, Havlicek

, Rachakonda

, Fries

, Kalyanam

, Michael

, Caprihan

, Turner

, Eichele

, Adelsheim

, Bryan

, Bustillo

, Clark

, Feldstein Ewing

, Filbey

, Ford

, Hutchison

, Jung

, Kiehl

, Kodituwakku

, Komesu

, Mayer

, Pearlson

, Phillips

, Sadek

, Stevens

, Teuscher

, Thoma

, Calhoun

(2011) A baseline for the multivariate comparison of resting-state networks. Front Syst Neurosci 5, 2.

28.

Deco

, Jirsa

, McIntosh

(2013) Resting brains never rest: computational insights into potential cognitive architectures. Trends Neurosci 36, 268–274.

29.

Deco

, Ponce-Alvarez

, Mantini

, Romani

, Hagmann

, Corbetta

(2013) Resting-state functional connectivity emerges from structurally and dynamically shaped slow linear fluctuations. J Neurosci 33, 11239–11252.

30.

Drzezga

, Becker

, Van Dijk

, Sreenivasan

, Talukdar

, Sullivan

, Schultz

, Sepulcre

, Putcha

, Greve

, Johnson

, Sperling

(2011) Neuronal dysfunction and disconnection of cortical hubs in non-demented subjects with elevated amyloid burden. Brain 134, 1635–1646.

31.

Cabeza

, Ciaramelli

, Olson

, Moscovitch

(2008) The parietal cortex and episodic memory: an attentional account. Nat Rev Neurosci 9, 613–625.

32.

Myers

, Pasquini

, Göttler

, Grimmer

, Koch

, Ortner

, Neitzel

, Mühlau

, Förster

, Kurz

, Förstl

, Zimmer

, Wohlschläger

, Riedl

, Drzezga

, Sorg

(2014) Within-patient correspondence of Amyloid beta and intrinsic network connectivity in Alzheimer’s disease. Brain 137, 2052–2064.

33.

Pasquini

, Benson

, Grothe

, Utz

, Myers

, Yakushev

, Grimmer

, Scherr

, Sorg

, Alzheimer’s Disease Neuroimaging Initiative (2017) Individual correspondence of amyloid beta and intrinsic connectivity in the posterior default mode network across stages of Alzheimer’s disease. J Alzheimers Dis 58, 763–773>.

34.

Bookheimer

, Strojwas

, Cohen

, Saunders

, Pericak-Vance

, Mazziotta

, Small

(2000) Patterns of brain activation in people at risk for Alzheimer’s disease. N Engl J Med 343, 450–456.

35.

Celone

, Calhoun

, Dickerson

, Atri

, Chua

, Miller

, DePeau

, Rentz

, Selkoe

, Blacker

, Albert

, Sperling

(2006) Alterations in memory networks in mild cognitive impairment and Alzheimer’s disease: an independent component analysis. J Neurosci 26, 10222–10231.

36.

Dickerson

, Salat

, Greve

, Chua

, Rand-Giovannetti

, Rentz

, Bertram

, Mullin

, Tanzi

, Blacker

, Albert

, Sperling

(2005) Increased hippocampal activation in mild cognitive impairment compared to normal aging and AD. Neurology 65, 404–411.

37.

Riedl

, Bienkowska

, Strobel

, Tahmasian

, Grimmer

, Förster

, Friston

, Sorg

, Drzezga

(2014) Local activity determines functional connectivity in the resting human brain: a simultaneous FDG-ET/fMRI study. J Neurosci 34, 6260–6266.