Research organism

elife

eLife

2050-084X

eLife Sciences Publications, Ltd

62114

10.7554/eLife.62114

Research Article

Neuroscience

Longitudinal stability of medial temporal lobe connectivity is associated with tau-related memory decline

Chen

Quanjing

https://orcid.org/0000-0003-4630-6817

quanjingchen@gmail.com12Turnbull

Adam

13Baran

Timothy M

34Lin

Feng V

1256781Elaine C. Hubbard Center for Nursing Research on Aging, School of Nursing, University of Rochester Medical Center

Rochester

United States2Department of Psychiatry, School of Medicine and Dentistry, University of Rochester Medical Center

Rochester

United States3Department of Imaging Sciences, School of Medicine and Dentistry, University of Rochester Medical Center

Rochester

United States4Department of Biomedical Engineering, University of Rochester

Rochester

United States5Department of Neuroscience, School of Medicine and Dentistry, University of Rochester Medical Center

Rochester

United States6Department of Neurology, School of Medicine and Dentistry, University of Rochester Medical Center

Rochester

United States7Department of Brain and Cognitive Sciences, University of Rochester

Rochester

United States8School of Medicine, Stanford University

Stanford

United States

Irish

Muireann

Reviewing EditorUniversity of SydneyAustraliaBaker

Chris I

Senior EditorNational Institute of Mental Health, National Institutes of HealthUnited States

31122020

2020

e62114

1408202030122020

2020

Chen et al

http://creativecommons.org/licenses/by/4.0/

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

The relationship between Alzheimer’s disease (AD) pathology and cognitive decline is an important topic in the aging research field. Recent studies suggest that memory deficits are more susceptible to phosphorylated tau (Ptau) than amyloid-beta. However, little is known regarding the neurocognitive mechanisms linking Ptau and memory-related decline. Here, we extracted data from Alzheimer’s Disease Neuroimaging Initiative (ADNI) participants with cerebrospinal fluid (CSF) Ptau collected at baseline, diffusion tensor imaging measure twice, 2 year apart, and longitudinal memory data over 5 years. We defined three age- and education-matched groups: Ptau negative cognitively unimpaired, Ptau positive cognitively unimpaired, and Ptau positive individuals with mild cognitive impairment. We found the presence of CSF Ptau at baseline was related to a loss of structural stability in medial temporal lobe connectivity in a way that matched proposed disease progression, and this loss of stability in connections known to be important for memory moderated the relationship between Ptau accumulation and memory decline.

structural connectivitymedial temporal lobealzheimer's diseasephosphorylated tau

Research organismHuman

http://dx.doi.org/10.13039/100000002

National Institutes of Health

R01 NR015452

Lin

Feng V

http://dx.doi.org/10.13039/100000002

National Institutes of Health

U01 AG024904

Lin

Feng V

http://dx.doi.org/10.13039/100000005

Department of Defense

W81XWH-12-2-0012

Lin

Feng V

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Author impact statement

Phosphorylated tau was related to a loss of structural stability in medial temporal lobe connectivity, and this loss of stability moderated the relationship between phosphorylated tau accumulation and memory decline.

Introduction

Alzheimer’s disease (AD) is characterized by the presence of both amyloid-beta (Aβ) and tau proteins that lead to neurodegeneration (Jack et al., 2015). The dominant view in the literature is that Aβ pathology triggers tau aggregation, strongly supported by evidence from autosomal dominant AD (Hardy, 2002). However, there have been important recent findings from sporadic AD that suggest that hyperphosphorylated tau proteins that are no longer governed by normal cellular removal mechanisms play a necessary and underappreciated role in AD progression (Musiek and Holtzman, 2015), with some authors even suggesting that this may be the primary step in the disease needed for subsequent AB aggregation (Arnsten et al., 2020; Braak and Del Tredici, 2015). In addition, a recent study found that tau-positron emission tomography (PET), but not β-amyloid–PET, could predict longitudinal brain atrophy in AD (La Joie et al., 2020). Perhaps for this reason, across animal and human studies, phosphorylated tau (Ptau), rather than Aβ, seems to better reflect cognitive deficits (Giannakopoulos et al., 2003; Hedden et al., 2013; Huber et al., 2018; Mitchell et al., 2002; Bennett et al., 2004; Schöll et al., 2016). However, for Ptau to become a potentially meaningful therapeutic target for preventing or slowing dementia progression, it is essential to better understand the mechanisms linking Ptau and cognitive deficits.

Both in vivo and in vitro findings support that abnormal tau appears to originate in the anterior medial temporal lobe (MTL) before spreading to anterior-temporal (AT) and posterior-medial (PM) regions of the cerebral cortex (Braak and Del Tredici, 2015; Cho et al., 2016). Cumulative evidence suggests that AD is a connectome disease (Acosta-Cabronero et al., 2010, Dai and He, 2014, Damoiseaux et al., 2009, Delbeuck et al., 2003, Rose et al., 2000), and abnormal tau spreads via neuronal connections (Clavaguera et al., 2009). MTL is strongly connected with both the AT and PM systems, with anterolateral entorhinal cortex showing preferential connectivity with anterior regions overlapping with the AT system and parahippocampal gyrus and posteromedial entorhinal cortex with posterior regions overlapping with the PM system (Maass et al., 2015; Navarro Schröder et al., 2015). The accumulation of abnormal tau in the MTL and subsequent spread to these cortical memory systems are likely to be particularly relevant for understanding memory deficits in AD, as these systems play dissociable roles in human memory: the AT system is specific for object-related stimuli (i.e., semantic memory) and the PM is specific for remembering scenes (Davachi et al., 2003; Epstein and Kanwisher, 1998; Wang et al., 2020). Taken together, these findings suggest that damage to the integrity of neural connections between the MTL, AT, and PM may be a mechanism by which Ptau accumulation leads to cognitive decline.

Several studies have examined the relationship between the functional (Berron et al., 2020; Cope et al., 2018; Franzmeier et al., 2020; Franzmeier et al., 2019; Maass et al., 2015; Ossenkoppele et al., 2019) and structural connectivity and tau pathology (Jacobs et al., 2018; Kim et al., 2019). Tau propagation from the entorhinal cortex to anterior and posterior brain regions that overlap with the AT and PM memory systems has been shown to relate to connectivity strength as measured by functional connectivity in cognitively normal adults, consistent with the idea that tau spreads via neuronal connections (Adams et al., 2019). Furthermore, tau pathology is related to worse performance in object memory tasks involving the AT system relative to scene memory tasks involving the PM system in cognitive normal older adults, also matching the proposed spread of the disease via neuronal connections, as tau burden was much higher in the AT system compared with the PM system in the early stages of AD (Arnsten et al., 2020; Maass et al., 2019). While single measures of functional and structural connectivity can give clues into the spread of Ptau, abnormalities in the longitudinal integrity of the structural connectivity may directly reflect loss of axonal integrity from tau pathology (Conturo et al., 1999). Especially during predementia AD stages when AD pathology is relatively mild, the loss of this integrity due to tau pathology might help explain how tau accumulation in the MTL leads to memory deficits. Synthesizing these separate lines of evidence, we suspect that the longitudinal integrity of the structural connections of the MTL may link Ptau accumulation and episodic memory decline in predementia AD syndromes.

Here, we capture the longitudinal integrity of the structural connectivity by calculating correlations across multiple time points to give a measure of stability. This novel concept has been first applied to functional connectivity (Kaufmann et al., 2018; Kaufmann et al., 2017). An emerging study suggests that a functional connectome stability measure is associated with cognition (average and rate of episodic memory decline) (Ousdal et al., 2020). Traditional univariate tract-wise analysis first derives a change score of each tract and then compares it across participants. Conversely, the stability measure calculates the pattern similarity between two time points across multiple connections within participants, in which subtle changes of individual connections may affect the stability score. The stability measure may therefore be suitable to capture subtle structural changes in the predementia AD stage.

We aimed to examine the relationship between structural stability, Ptau, and episodic memory. We extracted data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database, focusing on participants with cerebrospinal fluid (CSF) Ptau collected at baseline, as well as diffusion tensor imaging (DTI) measure twice, 2 years apart with the baseline measure corresponding to Ptau measurement. In addition, we extracted episodic memory data at baseline (anchored by Ptau measurement) and throughout four additional years.

We had several hypotheses: first, based on the idea that the accumulation of Ptau in MTL in the predementia AD stages causes a loss of integrity in the axons that connect MTL to AT and PM systems, we hypothesized that higher CSF Ptau levels will be associated with decreased MTL-related structural stability. Specifically, we examined stability within the MTL, as well as in MTL-AT and MTL-PM connections. Second, given that tau tangles deposit early in the AT lobe (Arnsten et al., 2020; Braak and Del Tredici, 2015; Maass et al., 2019), we hypothesized that Ptau first affects the stability in the MTL and MTL-AT in the predementia stages, followed by MTL-PM as the disease progresses and tau pathology worsens and spreads to posterior brain regions. Finally, given that MTL and related structures play a central role in episodic memory, we hypothesized that MTL-related structural stability might be associated with tau-related episodic memory decline.

Results

We extracted data from ADNI participants with CSF Ptau collected at baseline, DTI measure twice, 2 year apart, and longitudinal episodic memory data over 5 years. We defined three age- and education-matched groups based on baseline Ptau and cognitive impairment: Ptau negative cognitively unimpaired (CN Ptau–, n = 26), Ptau positive cognitively unimpaired (CN Ptau+, n = 18), and Ptau positive individuals with mild cognitive impairment (MCI Ptau+, n = 30). Group comparison of baseline characteristics is presented in Table 1. There was no significant difference between the CN Ptau– and CN Ptau+ groups, except for the Ptau pathology (p < 0.001). Compared with CN groups, MCI Ptau+ group had more APOEε4 carriers, higher Ptau levels, lower CSF Aβ level, greater neurodegeneration, as well as worse performance in episodic memory and global cognition (all ps < 0.05).

Table 1.Baseline characteristics.

Table 1—source data 1.Baseline characteristics.

	CN Ptau– (n = 26)	CN Ptau+ (n = 18)	MCI Ptau+ (n = 30)	T, F, or χ² tests, df1, df2, (p)
Age baseline, mean (SD)	74.0 (4.63)	74.8 (6.04)	73.9 (7.34)	0.14, 2, 71 (0.87)
Age baseline ≥ 75, N (%)	12 (46.2)	8 (44.4)	15 (50.0)	0.16, 2 (0.92)
Male, N (%)	11 (42.3)	7 (38.9)	18 (60.0)	2.65, 2 (0.27)
Education, mean (SD)	16.69 (2.57)	17.28 (2.42)	15.67 (2.35)	2.68, 2, 71 (0.08)
APOEε4 carrier, N (%)	5 (19.2)^a	7 (38.8)^a	22 (73.3)^b	14.79, 2 (<0.001)
CSF Ptau baseline, mean (SD)	15.8 (4.30)^a	29.0 (7.63)^b	38.0 (12.34)^c	41.58, 2, 71 (<0.001)
CSF Aβ baseline, mean (SD)	1231.1 (571.89)^a	1396.2 (690.95)^a	890.7 (383.11)^b	5.63, 2, 71 (0.005)
CSF Aβ-positive baseline, N (%)	9 (34.6)^a	7 (38.9)^a	22 (73.3)^b	9.84, 2 (0.007)
Neurodegeneration baseline, mean (SD)	2.97 (0.11)^a	2.89 (0.17)^a	2.77 (0.20)^b	9.68, 2, 71 (<0.001)
Neurodegeneration-positive baseline, N (%)	2 (7.69)^a	3 (16.67)^a	14 (46.67)^b	12.10, 2 (0.002)
Episodic memory baseline, mean (SD)	0.75 (0.81)^a	0.67 (0.79)^a	0.14 (0.84)^b	4.60, 2, 71 (0.013)
MOCA baseline, mean (SD)	25.92 (2.07)^a	25.89 (2.54)^a	21.67 (2.17)^b	32.17, 2, 71 (<0.001)

Note: CN, cognitively normal; MCI, amnestic mild cognitive impairment; APOEε4, apolipoprotein E ε4; SD, standard deviation; CSF, cerebrospinal fluid; Aβ, beta-amyloid-(1–42); Ptau, phosphorylated tau; MOCA, Montreal Cognitive Assessment. a, b, c represents the post hoc comparison difference from the F-test or χ² test. Bold values indicate p < 0.05.

Longitudinal integrity of structural connections was measured as a stability index of correlations across baseline and 2 year follow-up for connections between the MTL and cortical memory systems (see Materials and methods for details). We identified regions of interests based on the Desikan-Killiany Atlas. Since the hippocampus and the surrounding hippocampal region including the parahippocampal cortex and entorhinal cortex are the primary regions supporting memory processing (Davachi et al., 2003), we consider them to belong to MTL memory network. Thus, the MTL system includes entorhinal, hippocampus, and parahippocampal gyrus in both hemispheres. In line with previous literature (Berron et al., 2020; Ranganath and Ritchey, 2012), the AT system includes bilateral inferior temporal, temporal polar, and lateral and medial orbitofrontal cortex, while the PM system includes bilateral posterior and isthmus cingulate, lateral occipital cortex, precuneus, and thalamus. Figure 1A shows the center of mass of regions of interests. For visualization purposes, we presented the connections generated between MTL-related structures in a representative participant in Figure 1B. The connections within MTL include the hippocampal cingulum bundle and fornix. The connections between MTL and AT largely overlap with the anterior segments of the inferior longitudinal fasciculus (ILF) and inferior fronto-occipital fasciculus (IFOF), while the connections between MTL and PM mainly involve the cingulum bundle and posterior segments of ILF and IFOF.

Figure 1.Regions of interests.

(A) Regions of interests are identified based on the Desikan-Killiany Atlas and visualized with the BrainNet Viewer (Xia et al., 2013). The medial temporal lobe (MTL) includes bilateral entorhinal cortex (ENT), hippocampus (HIP), and parahippocampal gyrus (PHIP). The anterior-temporal (AT) system includes bilateral inferior temporal cortex (IT), temporal pole (TP), and lateral and medial orbitofrontal cortex (LOF/MOF), while the posterior-medial (PM) system includes bilateral posterior and isthmus cingulate (PCC/ISTC), lateral occipital cortex (LO), precuneus (PREC), and thalamus (TH). (B) Visualization of connections between MTL-related structures in a representative participant with TractVis (Wang and Wedeen, 2007).

Figure 1—figure supplement 1.Average mean diffusivity (MD) matrices for each group at each time point.

Please note that the white color indicates that no streamline was generated between the region of interest pairs in any of the participants within the group.

Figure 1—figure supplement 1—source data 1.Average mean diffusivity (MD) matrices for each group at each time point.

Relationship between baseline CSF Ptau and MTL structural stability

We hypothesized that Ptau affects the stability in MTL and MTL-AT early in predementia AD stages (e.g., CN Ptau+) and with disease progression and increased tau pathology (e.g., MCI Ptau+), the stability in MTL-PM is also disrupted. To test this, one-tailed independent t-tests were conducted between (1) CN Ptau+ and CN Ptau–; (2) MCI Ptau+ and CN Ptau–; and (3) MCI Ptau+ and CN Ptau+. We expected to see disrupted stability in MTL and MTL-AT in CN Ptau+, compared with CN Ptau–. For MCI Ptau+, the disruption would extend to MTL-PM, compared to both CN Ptau– and CN Ptau+. One-tailed tests were used because our hypotheses were directional, expecting worse stability with disease progression.

In line with our hypothesis that MTL and MTL-AT stability is affected by Ptau early in the predementia AD stage, compared to CN Ptau–, structural stability was significantly lower in CN Ptau+ in MTL (t(42) = –2.11, p = 0.022) and MTL-AT (t(42) = –2.57, p = 0.007). Decreased structural stability was further found in MTL (t(54) = –2.61, p = 0.006) and MTL-AT (t(54) = –3.55, p < 0.001) in MCI Ptau+, compared to CN Ptau-. In line with our hypothesis that MTL-PM is affected later in disease progression, we found a significantly lower stability in MTL-PM in MCI Ptau+, compared to CN Ptau+ (t(46) = –5.12, p < 0.001) and CN Ptau– (t(54) = –3.89, p < 0.001, Figure 2A). These results suggested that Ptau affects the structural integrity of connections from the MTL in a way that follows the proposed spread of the disease via neuronal connections.

Figure 2.Relationship between baseline CSF Ptau and MTL structural stability.

(A) Structural stability within MTL, MTL-AT, and MTL-PM for CN Ptau– (blue), CN Ptau+ (orange), and MCI Ptau+ (red) groups. Asterisk represents significant group comparison. *p < 0.05; **p < 0.01; ***p < 0.001. (B) Relationship of baseline CSF Ptau levels and structural stability within MTL, MTL-AT, and MTL-PM for the entire sample.

Figure 2—source data 1.Group comparisons.

Figure 2—figure supplement 1.Group comparisons without the exclusion of eight participants.

In addition, using the entire sample, Model 1 (Y stability = β10 + β11Ptau + ε₁) suggested that levels of Ptau significantly related to structural stability in MTL (B = –0.61, SE = 0.23, Wald’s χ² = 7.32, BH-adjusted p = 0.011, Figure 2B, left), MTL-AT (B = –0.59, SE = 0.21, Wald’s χ² = 7.73, BH-adjusted p = 0.011, Figure 2B, middle), and MTL-PM (B = –0.37, SE = 0.17, Wald’s χ²= 4.43, BH-adjusted p = 0.035, Figure 2B, right).

The effect of age, CSF Aβ pathology, and neurodegeneration on MTL structural stability in the whole sample

We examined whether structural stability could be directly affected by covariates (i.e., age, CSF Aβ, or neurodegeneration) in the entire sample with Model 2 (Y stability = β20 + β21Covariates + ε₂). We found no significant relationship between age and stability (all ps > 0.05, Figure 3A). We found a significant correlation between CSF Aβ and stability in MTL-PM (β = 0.40, SE = 0.16, Wald’s χ² = 6.04, BH-adjusted p = 0.042, Figure 3B). Neurodegeneration significantly related to stability within MTL (β = 0.58, SE = 0.26, Wald’s χ² = 4.97, BH-adjusted p = 0.039), MTL-AT (β = 0.49, SE = 0.25, Wald’s χ² = 3.89, BH-adjusted p = 0.048), and MTL-PM (β = 0.61, SE = 0.19, Wald’s χ² = 10.39, BH-adjusted p = 0.003, Figure 3C).

Figure 3.The effect of age (<bold>A</bold>), CSF Aβ pathology (<bold>B</bold>), and neurodegeneration (<bold>C</bold>) on the structural stability within MTL, MTL-AT, and MTL-PM in the whole sample.

N: neurodegeneration; higher values indicate greater cortical thickness and lower severity in neurodegeneration.

Figure 3—source data 1.The effect of covariates on the structural stability in the whole sample.

The effect of age, CSF Aβ pathology, and neurodegeneration on the relationship between Ptau and the MTL structural stability in the whole sample

We further examined whether the relationship between Ptau and stability would be affected by covariates (i.e., age, Aβ, or neurodegeneration). We tested Ptau and each covariate’s interaction effect, controlling for their main effects, in predicting the structural stability in the whole sample with Model 3 (Y stability = β30 + β31Ptau + β32Covariate + β33Ptau× Covariates + ε₃). Results showed no significant interaction for Ptau × age, Ptau × CSF Aβ, or Ptau × neurodegeneration (Figure 4), suggesting that the relationship between Ptau and stability is not affected by age, CSF Aβ pathology, or neurodegeneration.

Figure 4.The effect of age (<bold>A</bold>), CSF Aβ pathology (<bold>B</bold>), and neurodegeneration (<bold>C</bold>) on the relationship between baseline CSF Ptau levels and the structural stability within MTL, MTL-AT, and MTL-PM in the whole sample.

Figure 4—source data 1.The effect of covariates on the relationship between Ptau and structural stability.

Relationship between MTL structural stability, Ptau, and memory over 5 years in the whole sample

Having determined that Ptau related to our measure of structural stability in a way that suggests it follows proposed models of disease progression, we next assessed whether structural stability predicts episodic memory. In line with our hypothesis that decreased stability in MTL-related structures leads to deficits in episodic memory, results from Model 4 (Y average episodic memory = β40 + β41 Stability + ε₄) suggested that stability in MTL (B = 0.60, SE = 0.24, Wald’s χ² = 6.23, BH-adjusted p = 0.020, Figure 5A, left), MTL-AT (B = 0.56, SE = 0.26, Wald’s χ² = 4.73, BH-adjusted p = 0.030, Figure 5A, middle), and MTL-PM (B = 0.95, SE = 0.31, Wald’s χ² = 9.29, BH-adjusted p = 0.006, Figure 5A, right) significantly predicted average episodic memory across 5 years. When we included the covariates (i.e., age, education, sex, neurodegeneration, and CSF Aβ), all results remained significant (MTL: B = 0.42, SE = 0.21, Wald’s χ² = 3.98, BH-adjusted p = 0.046; MTL-AT: B = 0.47, SE = 0.22, Wald’s χ² = 4.52, BH-adjusted p = 0.046; MTL-PM: B = 0.88, SE = 0.29, Wald’s χ² = 9.10, BH-adjusted p = 0.009). Furthermore, results from Model 4’ (Y rate of episodic memory change = β40’ + β41’ Stability + ε_4’) showed lower stability in MTL (B = 0.12, SE = 0.06, Wald’s χ² = 4.13, BH-adjusted p = 0.040, Figure 5B, left) and MTL-PM (B = 0.15, SE = 0.07, Wald’s χ² = 4.04, BH-adjusted p = 0.042, Figure 5B, right) related to greater decreased rate of episodic memory. We did not find any relationship between MTL-AT stability and rate of episodic memory change (Figure 5B, middle). Only the result in MTL remained significant after the covariates were controlled (B = 0.11, SE = 0.06, Wald’s χ² = 3.85, p = 0.050).

Figure 5.MTL structural stability predicts episodic memory.

(A) Relationship of average episodic memory over 5 years and the structural stability in the whole sample. (B) Relationship of episodic memory decline rate and the structural stability in the whole sample.

Figure 5—source data 1.MTL structural stability predicts episodic memory.

To this point, our analyses have established that levels of Ptau significantly related to structural integrity in a manner that follows proposed disease progression, with Ptau in patients early in the disease relating to a loss of structural stability in MTL and MTL-AT connections, followed by MTL-PM connections over time as the disease progresses to MCI. We have also shown that decreased stability in the MTL, MTL-AT, and MTL-PM connections related to worse average episodic memory. Specifically, decreased stability within the MTL connections predicted future rate of memory decline. Our final analysis sought to formally test the hypothesis that this loss of structural stability in MTL provides a link between Ptau accumulation and memory decline. We expected to see stability as a significant moderator of the relationship between Ptau and episodic memory.

Results from Model 5 (Y average episodic memory = β50 + β51 Ptau + ε₅) suggested that Ptau at baseline predicted average episodic memory across 5 years (B = –1.09, SE = 0.49, Wald’s χ² = 4.85, p = 0.028). After adding stability and the interaction between stability and Ptau (Model 6: Y average episodic memory = β60+ β61Ptau + β62Stability + β63Ptau × Stability + ε₆), there was no significant interaction between Ptau and stability of MTL (B = 1.34, SE = 1.62, Wald’s χ² = 0.69, p = 0.41).

Results from Model 5’ (Y rate of episodic memory change = β50’ + β51’ Ptau + ε_5’) suggested that higher level of Ptau predicted greater decreased rate of memory (B = –0.38, SE = 0.11, Wald’s χ² = 11.34, p = 0.001). After adding stability and the interaction between stability and Ptau into the model (Model 6’: Y rate of episodic memory change = β60_’ + β61_’Ptau + β62_’Stability + β63_’Ptau × Stability + ε_6’), Ptau × Stability of MTL was significant (B = –0.01, SE = 0.01, Wald’s χ² = 5.28, p = 0.022). The interaction effect remained significant while accounting for covariates (B = –0.01, SE = 0.01, Wald’s χ² = 3.97, p = 0.046). Figure 6 displayed the associations between Ptau and rate of memory change depending on MTL stability. Participants were divided based on terciles of MTL stability. Stronger relationship between Ptau and rate of memory change was observed in those with higher MTL stability. All of these results suggested that the structural stability of the MTL links Ptau and episodic memory.

Figure 6.The moderating effect of medial temporal lobe (MTL) structural stability on the relationship between Ptau and rate of memory change.

The plot shows how structural stability within MTL affected the relationship between Ptau and rate of memory change. Participants were divided into terciles based on their MTL stability. The lowest tercile is shown in red, the middle in green, and the highest in blue. P-values indicate whether the slope of the regression line is significantly different from zero for each tercile separately.

Figure 6—source data 1.MTL structural stability moderates the relationship between Ptau and memory change.

Discussion

In this study, we examined whether the presence of CSF Ptau at baseline was related to a loss of structural stability in a way that matched proposed disease progression and whether this loss of stability in connections important for episodic memory is related to the link between Ptau accumulation and cognitive decline. We did this by first examining the relationship between Ptau at baseline and longitudinal white matter stability across three groups (CN Ptau–, CN Ptau+, and MCI Ptau+). CN Ptau+ and MCI Ptau+ had decreased stability in MTL and MTL-AT compared to CN Ptau-, while MCI Ptau+ further showed decrease in MTL-PM stability relative to both CN Ptau– and CN Ptau+ groups, suggesting that Ptau relates to structural stability in a temporal pattern that matches its proposed spread from the MTL via neuronal connections. Furthermore, the stability in MTL, MTL-AT, and MTL-PM predicted average episodic memory over 5 years, suggesting that damage to these pathways specifically relates to episodic memory decline. Importantly, structural stability in MTL moderated the effect of Ptau on the rate of memory change, suggesting that structural stability in MTL may be a candidate as a mechanistic link between the accumulation of Ptau and cognitive decline.

Ptau influences the white matter structural stability

Comparing groups with varied degrees of Ptau abnormality (Ptau+ vs. Ptau–) and/or cognitive deficits (CN vs. MCI), MTL structural stability was disrupted both within MTL and between MTL and AT in CN Ptau+ and MCI Ptau+ relative to CN Ptau–, and the disruption was further extended to MTL-PM connections in MCI Ptau+ relative to both CN Ptau– and CN Ptau+. This is consistent with the proposed model of Ptau spread from the MTL via neuronal connections. In a recent longitudinal study examining Ptau’s influence on the structural connectome in the preclinical stage of AD, elevated Ptau was associated with changes primarily in MTL and MTL-AT regions (Kim et al., 2019). Another study found that Ptau accumulation in posterior cingulate cortex (a region of PM) could be predicted by baseline mean diffusivity (MD) in MTL (Jacobs et al., 2018). Noticeably in our study, the relationship between Ptau and structural stability was robust, as it was barely affected by neurodegeneration, Aβ, or age. In summary, our results add to previous literature suggesting that tau pathology may act as the primary mechanism for changes in white matter structural integrity.

White matter structural stability predicts longitudinal episodic memory

In line with previous findings (Jacobs et al., 2018), our study revealed a robust relationship between structural stability in MTL, MTL-AT, and MTL-PM connections and average episodic memory. Additionally, stability of MTL related to the decline rate of memory over time. Literature suggests that the decline in episodic memory can be predicted by changes in functional connectivity between select regions (e.g., cortico-striatal [Fjell et al., 2016], subcortical [Ousdal et al., 2020], default mode networks [Staffaroni et al., 2018], and posterior medial temporal lobe [Salami et al., 2016]). Our findings add to this literature by showing that an index reflecting the longitudinal integrity of MTL connections relates to episodic memory decline over time, suggesting that changes in structural integrity of within MTL connections over the course of several years co-occur with episodic memory decline. These results suggested that MTL-related structural connections might serve as potential therapeutic targets for preventing memory decline in the early stage of AD.

MTL structural stability links Ptau and longitudinal episodic memory

Previous research has shown that MTL structural connectivity influences downstream PM Ptau, and changes in both MTL white matter and PM Ptau influenced memory (Jacobs et al., 2018). However, it is unclear whether changes in MTL structural connectivity are a result of accumulated Ptau and whether these changes present a mechanism by which Ptau may lead to decreases in episodic memory. In the current study, we found that levels of Ptau were related to decreases in MTL structural stability and that these changes were in turn related to decreases in episodic memory. A model showed that decreases in the stability of within MTL connections significantly predicted episodic memory decline, suggesting that changes in these connections resulting from increased Ptau are associated with decreases in episodic memory. Synthesizing the relationship between Ptau and structural stability, as well as structural stability and episodic memory, we found MTL structural stability moderated the relationship between Ptau and rate of memory change. Ptau’s predictive strength for cognition is overall limited, which interferes with its clinical value for early detection of cognitive decline (Hansson and Mormino, 2018). Structural stability in MTL may be a viable biomarker and early intervention target addressing the linkage between pathology and cognition. It is possible that the observation of a stronger relationship between more Ptau and faster rate of memory decline in those with higher MTL stability might be driven by greater variance in this group, caused by the inclusion of both stable individuals with overall low connectivity as well as consistently high connectivity. To assess this, we looked at the number of spread of individuals across these groups based on their mean structural connectivity in MTL (Table 2, see Materials and methods for details). The stable group contains 12 high–high MD and 15 low–low MD subjects, while the instable group contains 12 high–high MD and nine low–low MD subjects. While there are more low-low subjects in the stable group, the ratio of low-low to high–high is similar across these groups, suggesting increased variance is unlikely to drive the strong relationship with cognition in the high-stability group.

Table 2.Dichotomized mean MD and stability of MTL.

Table 2—source data 1.Dichotomized mean MD and stability of MTL.

Mean MD of MTL at year 0	Mean MD of MTL at year 2	Stability of MTL	CN Ptau– (N = 26)	CN Ptau+ (N = 18)	MCI Ptau+ (N = 30)
High	High	Stable	8 (30.8%)	0 (0%)	4 (13.3%)
High	Low	Stable	0 (0%)	0 (0%)	3 (10.0%)
Low	High	Stable	6 (23.1%)	1 (5.6%)	0 (0%)
Low	Low	Stable	4 (15.4%)	7 (38.9%)	4 (13.3%)
High	High	Instable	2 (7.6%)	6 (33.3%)	4 (13.3%)
High	Low	Instable	0 (0%)	2 (11.1%)	8 (26.7%)
Low	High	Instable	3 (11.5%)	0 (0%)	3 (10.0%)
Low	Low	Instable	3 (11.5%)	2 (11.1%)	4 (13.3%)

Note: Mean MD is the averaged MD across all connections within MTL. MD, mean diffusivity; MTL, medial temporal lobe.

Limitations

This study has several limitations. The major limitation of this study was related to its small sample size, since we have only 18 participants in the CN Ptau+ group. In future studies with larger sample size, data-driven analyses could be conducted to validate our findings and explore other relevant findings. Second, Ptau is determined using one-time measurement from CSF rather than longitudinal PET. All regionally specific hypotheses on tau accumulation are based on previous studies, rather than observations in the current study. Finally, the 2 year interval for DTI data was relatively short, which may not be sufficient to capture reliable changes in structural connectivity in predementia AD. Longer follow-up period may reveal the significant interaction effect between Ptau and covariates (e.g., Aβ) in predicting the structural stability.

Materials and methodsData source

Data were obtained from ADNI GO and ADNI2 database (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), PET, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. For up-to-date information, see http://www.adni-info.org. ADNI GO and ADNI2 collect diffusion-weighted images (DWI) at 14 sites across North America, all using the same scanner manufacturer (General Electric), magnetic field strength (3T) protocol, including the same voxel size and gradient directions. To reduce site-related errors, ADNI group made rigor standardization and quality control over the protocols for patient recruitment and imaging data acquisition across sites. Each exam underwent a quality control evaluation at the Mayo Clinic (Rochester, MN), including inspection of each incoming image file for protocol compliance, clinically significant medical abnormalities, and image quality (Jack et al., 2018). ADNI3 was not used here due to different scan protocols and limited number of eligible participants at the time of manuscript preparation.

Participants

All participants in this study were from ADNI GO and ADNI2 where DTI data were collected using 3T GE scanners. Only non-demented individuals (healthy controls and MCI) with CSF pTau data at baseline and DTI data at two time points (2 years apart) were selected for this study. Participants were classified into different groups according to their CSF Ptau status and their clinical diagnosis, which resulted in the following participants: 29 Ptau negative cognitively unimpaired (CN Ptau–), 20 Ptau positive cognitively unimpaired (CN Ptau+), and 36 Ptau positive with MCI (MCI Ptau+). We then excluded three participants from the imaging data analysis due to inadequate tractography quality in either time of DTI data (none from CN Ptau–, one from CN Ptau+, and two from MCI), and eight participants to optimize age and education matching across groups (three participants aged <65 from CN Ptau–, one at age 89 from CN Ptau+, and four with years of education ≤ 12 from MCI Ptau+). The final sample contained 74 participants (26 CN Ptau–, 18 CN Ptau+, and 30 MCI Ptau+). Participant characteristics were provided in Table 1. The main result remains the same without the exclusion of eight participants (see Figure 2—figure supplement 1).

MeasuresCSF biomarkers

Ptau data were derived from CSF aliquots and measured using Elecsys CSF immunoassays at the University of Pennsylvania. The data is available in the ‘UPENNBIOMK10.csv’ file in the ADNI database. The cut-off for positive Ptau pathology was >21.8 pg/ml (Suárez-Calvet et al., 2019). Of note, Ptau data did not follow a normal distribution and were hence log10-transformed when used as a continuous variable.

Imaging dataAcquisition

All participants underwent whole-brain MRI scanning on 3T GE scanners. DWI were collected with the following parameters: matrix size = 256 × 256 mm; flip angle = 90°; slice thickness = 2.7 mm; 41 diffusion-weighted images (b = 1000 s/mm²) and five non-diffusion-weighted b0 image. T1-weighted spoiled gradient recalled echo sequences were acquired in the same scanning session (TR = 7.0–7.7 ms, TE = 2.8–3.2 ms, TI = 400 ms, matrix size = 256 × 256 mm, flip angle = 11°, resolution 1.02 × 1.02 mm, slice thickness = 1.20 mm). More details on ADNI protocols may be found at http://adni.loni.usc.edu/methods/documents/mri-protocols/.

Preprocessing

Each raw DWI image was aligned to the average b0 image using the FSL eddy_correct tool 6.0.1 (http://www.fmrib.ox.ac.uk/fsl) to correct for head motion and eddy current distortions. Non-brain tissue was removed using FSL’s Brain Extraction Tool (Smith, 2002). We then registered DWI images with the T1 anatomical images using Advanced Normalization Tools (ANTS; http://www.picsl.upenn.edu/ANTS/).

Network construction

To extract brain connectome, we employed an established structural connectome processing pipeline (for details, see Zhang et al., 2018). First, we applied a reproducible probabilistic tractography algorithm (Girard et al., 2014) to diffusion MRI data to generate streamlines across the whole brain. We used 0.2 mm step size and 25° angle threshold. Next, we defined the cortical, subcortical, and brainstem regions on T1 anatomical images for each participant using FreeSurfer (http://surfer.nmr.mgh.harvard.edu/). From the FreeSurfer segmentation, 34 regions per hemisphere were identified using Desikan-Killiany atlas (Desikan et al., 2006), as well as 17 subcortical and brainstem structures: the brain stem, and bilateral segmentations of thalamus, caudate, putamen, pallidum, hippocampus, amygdala, accumbens, and cerebellum. Together, we obtained 85 nodes per participant. For each pair of nodes, we extracted the streamlines connecting them. Next, we calculated the mean of mean diffusivity (MD), an indicator of myelination and axonal thickness, along each streamline to describe the connectivity (edge) between the connected nodes. This procedure generated an 85-by-85 symmetric MD-based structural connectivity matrix per participant at both baseline (year 0) and 2 year follow-up (year 2). Of note, previous studies have shown that MD is more sensitive than other metrics such as fractional anisotropy (FA) to white matter alterations associated with MCI (Yu et al., 2017) and AD (Jin et al., 2017), with MD also predicting conversion to dementia (Fellgiebel et al., 2006; Müller et al., 2007). Physiologically, the absolute diffusion, which is quantified by MD, is a more sensitive marker of neurodegeneration than FA, which simply quantifies the anisotropy of the diffusion tensor (Acosta-Cabronero et al., 2010). These findings motivated us to choose MD over other metrics.

Stability measure

For each individual, we computed longitudinal stability by calculating a within-participant Pearson correlation coefficients between year 0 and year 2 across connections. Each individual has stability scores for MTL, MTL-AT, and MTL-PM, respectively: the score for MTL is based on the 15 connections within MTL, MTL-AT based on the 48 unique connections between MTL and AT system, and MTL-PM based on 60 unique connections between MTL and PM system. Figure 1—figure supplement 1 shows the average mean diffusivity (MD) matrices for each group at year 0 and year 2, respectively.

The traditional univariate analysis only considers the overall magnitude of the network (e.g., the average MD across all connections within MTL network). By contrast, the stability measure is interested in the pattern of connectivity and reflects change in the patterns of MD across connections. This measure is purposefully meant to be agnostic to the mechanism by which stability is impacted across individuals; it is designed to capture changes in the pattern of connectivity that we believe occur in response to pathology. It may be that pathology causes connections to be reduced in strength, but other connections may show corresponding increases as a result. To understand how this measure related to overall connectivity, we averaged the MD measures across all connections within MTL to get a mean MD score for MTL network for each individual. Then we used the median as a cut-off to dichotomize the mean MD score to ‘high’ and ‘low’ for baseline (year 0) and 2 year follow-up (year 2), respectively. Similarly, we split stability of MTL to ‘stable’ and ‘instable’ at the median. The mean MD scores of 2 years and the MTL stability were dichotomized into two levels, over and under the cut point, forming eight groups in total. Table 2 displays the prevalence of eight groups in the CN/MCI samples. As we can see, it is possible that subjects maintained the same level of mean MD from baseline to follow-up (e.g., low–low or high–high), but their MD patterns actually changed and became instable. Thus, the traditional univariate analysis may not be as sensitive as the stability measure in detecting these subtle structural changes in the predementia AD stage.

Episodic memory measure

Episodic memory is a standardized composite score derived from multiple measures using factor analyses (Crane et al., 2012), on the memory-related domains of the Mini-Mental Status Examination, Alzheimer’s Disease Assessment Scale-Cognition subscale, Rey Auditory Verbal Learning Test, and Logical Memory Test. The average standardized episodic memory composite score is 0 among the entire ADNI sample (including those with cognitive impairment). We had follow-up data for episodic memory for all participants for up to 4 years (from baseline to a 4 year follow-up). To assess the association between longitudinal structural stability, baseline Ptau, and episodic memory, an average episodic memory score is calculated by averaging episodic memory over 5 years. Since we were interested in prediction of memory in the future, average memory score was used instead of baseline memory. Rate of memory change was determined by calculating the slope obtained from the linear regression line created by all available time points.

Covariates

We consider baseline demographics and factors associated with AD or episodic memory (i.e., age, Aβ, and neurodegeneration) as covariates in selected analysis. The cut-off for age was ≥75 years old since it has been widely used to classify young-old (65–74) and old-old (75+) in previous literature (Field and Minkler, 1988). Aβ were derived from CSF aliquots and measured using Elecsys CSF immunoassays at the University of Pennsylvania. The cut-off for positive Aβ was <976.6 pg/ml (Suárez-Calvet et al., 2019). A log-data transformation was applied to fit the skewed Aβ distribution into a normal distribution when used as a continuous variable. Neurodegeneration was quantified using structural MRI data by calculation of AD signature cortical thickness, consisting of posterior brain regions, including bilateral inferior and middle temporal lobes, entorhinal cortex, and fusiform gyrus, based on automated cortical parcellation using the Desikan-Killiany Atlas with Freesurfer (Jack et al., 2015; Lin et al., 2017). Higher values in AD signature cortical thickness indicate greater cortical thickness and lower severity in neurodegeneration. The cut-off for abnormal neurodegeneration was <2.77 mm (Jack et al., 2015).

Statistical analyses

To determine between-group differences in demographic and clinical information, one-way ANOVA and chi-square tests were used for continuous or categorical variables, respectively.

To examine the relationship between baseline CSF Ptau and structural stability, we conducted one-tailed independent t-tests between (1) CN Ptau+ and CN Ptau–; (2) MCI Ptau+ and CN Ptau–; and (3) MCI Ptau+ and CN Ptau+. In addition, we also conducted generalized linear model (GLM) using the entire sample with Ptau as a continuous variable.

Model 1: Y stability = β10 + β11Ptau + ε₁.

We examined whether structural stability could be directly affected by covariates (i.e., age, CSF Aβ or neurodegeneration) using GLMs.

Model 2: Y stability = β20 + β21Covariates + ε₂.

To examine whether the relationship between Ptau and stability would be affected by covariates (i.e., age, Aβ, or neurodegeneration), we tested Ptau and each covariate’s interaction effect, controlling for their main effects, in predicting the structural stability using a GLM in the whole sample.

Model 3: Y stability = β30 + β31Ptau + β32Covariate + β33Ptau× Covariates + ε₃.

We examined relationships between longitudinal structural stability, baseline Ptau, and episodic memory in the whole sample, including and excluding the covariates (age, education, sex, neurodegeneration, and CSF Aβ).

First, we expected to see that the stability in MTL-related structures could predict episodic memory. GLMs were conducted with and without covariates.

Model 4 (Stability predicts average episodic memory based on β31): Y average episodic memory = β40 + β41 Stability (+β42Covariates) + ε₄.

Model 4’ (Stability predicts decline rate of episodic memory based on β31’): Y rate of episodic memory change = β40’ + β41’ Stability (+β42’Covariates) + ε_4’.

Next, we tested whether Ptau could predict episodic memory using GLMs.

Model 5 (Ptau predicts average episodic memory based on β51): Y average episodic memory = β50 + β51 Ptau + ε₅.

Model 5’ (Ptau predicts decline rate of episodic memory over time based on β51’): Y rate of episodic memory change = β50’ + β51’ Ptau + ε_5’.

Finally, to examine whether stability moderated the relationship between Ptau and episodic memory, we tested the interaction between Ptau and Stability in model 6 or 6’ (β63 and β63’), including and excluding the covariates.

Model 6: Y average episodic memory = β60+ β61Ptau + β62Stability + β63Ptau× Stability (+β64Covariates) + ε₆.

Model 6’: Y rate of episodic memory change = β60’ + β61’Ptau + β62’Stability + β63’Ptau× Stability (+β64’Covariates) + ε_6’.

The analysis was conducted in SPSS 24.0 (IBM Corp., Armonk, NY). Results were corrected for multiple comparisons across MTL, MTL-AT, and MTL-PM using Benjamini–Hochberg (BH) procedure when appropriate and BH-adjusted p-values were reported.

Acknowledgements

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 and Development, LLC.; Johnson and Johnson Pharmaceutical Research and Development LLC.; Lumosity; Lundbeck; Merck and 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 (http://www.fnih.org). 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.

Additional information

Competing interests

No competing interests declared

Author contributions

Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft

Formal analysis, Writing - original draft, Writing - review and editing

Writing - original draft, Writing - review and editing

Conceptualization, Supervision, Funding acquisition, Writing - original draft, Writing - review and editing

Ethics

Human subjects: The current study is a secondary data analysis of limited-identified data per data user agreement between ADNI and F.L. The human subject research of original ADNI data collection was conducted at each ADNI data collection site (see the full list of sites http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf). Written informed consent was obtained from each participant (see http://adni.loni.usc.edu/wp-content/uploads/2008/07/adni2-procedures-manual.pdf for detailed information about ethical procedures for ADNI). Protocol_11.19.14.

Additional files

Transparent reporting form

Data availability

Data used in preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu/). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf.

The following previously published dataset was used:

JackCRBernsteinMAFoxNCThompsonPAlexanderGHarveyDBorowskiBBritsonPJWhitwellLJWardCDaleAM2008The Alzheimer's disease neuroimaging initiativeADNIadni

References

Acosta-Cabronero

Williams

Pengas

Nestor

2010

Absolute diffusivities define the landscape of white matter degeneration in Alzheimer's disease

Brain133529539

10.1093/brain/awp257

Adams

Maass

Harrison

Baker

Jagust

2019

Cortical tau deposition follows patterns of entorhinal functional connectivity in aging

eLife8

e49123

10.7554/eLife.49132

Arnsten

AFT

Datta

Tredici

Braak

2020

Hypothesis: tau pathology is an initiating factor in sporadic alzheimer's disease

Alzheimer's & Dementia5

12192

10.1002/alz.12192

Bennett

Schneider

Wilson

Bienias

Arnold

2004

Neurofibrillary tangles mediate the association of amyloid load with clinical alzheimer disease and level of cognitive function

Archives of Neurology61378384

10.1001/archneur.61.3.378

Berron

van Westen

Ossenkoppele

Strandberg

Hansson

2020

Medial temporal lobe connectivity and its associations with cognition in early Alzheimer's disease

Brain14312331248

10.1093/brain/awaa068

32252068

Braak

Del Tredici

2015

The preclinical phase of the pathological process underlying sporadic alzheimer's disease

Brain13828142833

10.1093/brain/awv236

26283673

Cho

Choi

Hwang

Kim

Lee

Ryu

Lee

Lyoo

2016

In vivo cortical spreading pattern of tau and amyloid in the alzheimer disease spectrum

Annals of Neurology80247258

10.1002/ana.24711

27323247

Clavaguera

Bolmont

Crowther

Abramowski

Frank

Probst

Fraser

Stalder

Beibel

Staufenbiel

Jucker

Goedert

Tolnay

2009

Transmission and spreading of tauopathy in transgenic mouse brain

Nature Cell Biology11909913

10.1038/ncb1901

19503072

Conturo

Lori

Cull

Akbudak

Snyder

Shimony

McKinstry

Burton

Raichle

1999

Tracking neuronal fiber pathways in the living human brain

PNAS961042210427

10.1073/pnas.96.18.10422

10468624

Cope

Rittman

Borchert

Jones

Vatansever

Allinson

Passamonti

Vazquez Rodriguez

Bevan-Jones

O'Brien

Rowe

2018

Tau burden and the functional connectome in Alzheimer's disease and progressive supranuclear palsy

Brain141550567

10.1093/brain/awx347

29293892

Crane

Carle

Gibbons

Insel

Mackin

Gross

Jones

Mukherjee

Curtis

Harvey

Weiner

Mungas

Alzheimer’s Disease Neuroimaging Initiative

2012

Development and assessment of a composite score for memory in the alzheimer's Disease Neuroimaging Initiative (ADNI)

Brain Imaging and Behavior6502516

10.1007/s11682-012-9186-z

22782295

Dai

2014

Disrupted structural and functional brain connectomes in mild cognitive impairment and Alzheimer’s disease

Neuroscience Bulletin30217232

10.1007/s12264-013-1421-0

Damoiseaux

Smith

Witter

Sanz-Arigita

Barkhof

Scheltens

Stam

Zarei

Rombouts

2009

White matter tract integrity in aging and Alzheimer's disease

Human Brain Mapping3010511059

10.1002/hbm.20563

18412132

Davachi

Mitchell

Wagner

2003

Multiple routes to memory: distinct medial temporal lobe processes build item and source memories

PNAS10021572162

10.1073/pnas.0337195100

12578977

Delbeuck

Van der Linden

Collette

2003

Alzheimer's disease as a disconnection syndrome?

Neuropsychology Review137992

10.1023/a:1023832305702

12887040

Desikan

Ségonne

Fischl

Quinn

Dickerson

Blacker

Buckner

Dale

Maguire

Hyman

Albert

Killiany

2006

An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest

NeuroImage31968980

10.1016/j.neuroimage.2006.01.021

16530430

Epstein

Kanwisher

1998

A cortical representation of the local visual environment

Nature392598601

10.1038/33402

9560155

Fellgiebel

Dellani

Greverus

Scheurich

Stoeter

Müller

2006

Predicting conversion to dementia in mild cognitive impairment by volumetric and diffusivity measurements of the Hippocampus

Psychiatry Research: Neuroimaging146283287

10.1016/j.pscychresns.2006.01.006

16530394

Field

Minkler

1988

Continuity and change in social support between young-old and old-old or very-old age

Journal of Gerontology43P100P106

10.1093/geronj/43.4.P100

3385147

Fjell

Sneve

Storsve

Grydeland

Yendiki

Walhovd

2016

Brain events underlying episodic memory changes in aging: a longitudinal investigation of structural and functional connectivity

Cerebral Cortex2612721286

10.1093/cercor/bhv102

25994960

Franzmeier

Rubinski

Neitzel

Kim

Damm

Kim

Lyoo

Cho

Finsterwalder

Duering

Seo

Ewers

Alzheimer’s Disease Neuroimaging Initiative

2019

Functional connectivity associated with tau levels in ageing, alzheimer's, and small vessel disease

Brain14210931107

10.1093/brain/awz026

30770704

Franzmeier

Neitzel

Rubinski

Smith

Strandberg

Ossenkoppele

Hansson

Ewers

Alzheimer’s Disease Neuroimaging Initiative (ADNI)

2020

Functional brain architecture is associated with the rate of tau accumulation in Alzheimer's disease

Nature Communications11

347

10.1038/s41467-019-14159-1

31953405

Giannakopoulos

Herrmann

Bussière

Bouras

Kövari

Perl

Morrison

Gold

Hof

2003

Tangle and neuron numbers, but not amyloid load, predict cognitive status in Alzheimer's disease

Neurology6014951500

10.1212/01.WNL.0000063311.58879.01

12743238

Girard

Whittingstall

Deriche

Descoteaux

2014

Towards quantitative connectivity analysis: reducing tractography biases

NeuroImage98266278

10.1016/j.neuroimage.2014.04.074

24816531

Hansson

Mormino

2018

Is longitudinal tau PET ready for use in Alzheimer’s disease clinical trials?

Brain14112411244

10.1093/brain/awy065

Hardy

2002

The amyloid hypothesis of Alzheimer's Disease: Progress and Problems on the Road to Therapeutics

Science297353356

10.1126/science.1072994

Hedden

Younger

Patel

2013

Meta-analysis of amyloid-cognition relations in cognitively normal older adults

Neurology8013411348

10.1212/WNL.0b013e31828ab35d

Huber

Yee

May

Dhanala

Mitchell

2018

Cognitive decline in preclinical alzheimer's Disease: Amyloid-Beta versus Tauopathy

Journal of Alzheimer's Disease61265281

10.3233/JAD-170490

29154274

Jack

Wiste

Weigand

Knopman

Mielke

Vemuri

Lowe

Senjem

Gunter

Reyes

Machulda

Roberts

Petersen

2015

Different definitions of neurodegeneration produce similar amyloid/neurodegeneration biomarker group findings

Brain13837473759

10.1093/brain/awv283

26428666

Jack

Bennett

Blennow

Carrillo

Dunn

Haeberlein

Holtzman

Jagust

Jessen

Karlawish

Liu

Molinuevo

Montine

Phelps

Rankin

Rowe

Scheltens

Siemers

Snyder

Sperling

Contributors

2018

NIA-AA research framework: toward a biological definition of alzheimer's disease

Alzheimer's & Dementia14535562

10.1016/j.jalz.2018.02.018

29653606

Jacobs

HIL

Hedden

Schultz

Sepulcre

Perea

Amariglio

Papp

Rentz

Sperling

Johnson

2018

Structural tract alterations predict downstream tau accumulation in amyloid-positive older individuals

Nature Neuroscience21424431

10.1038/s41593-018-0070-z

29403032

Jin

Huang

Daianu

Zhan

Dennis

Reid

Jack

Zhu

Thompson

Alzheimer's Disease Neuroimaging Initiative

2017

3d tract-specific local and global analysis of white matter integrity in Alzheimer's disease

Human Brain Mapping3811911207

10.1002/hbm.23448

27883250

Kaufmann

Alnæs

Doan

Brandt

Andreassen

Westlye

2017

Delayed stabilization and individualization in connectome development are related to psychiatric disorders

Nature Neuroscience20513515

10.1038/nn.4511

28218917

Kaufmann

Alnæs

Brandt

Bettella

Djurovic

Andreassen

Westlye

2018

Stability of the brain functional connectome fingerprint in individuals with schizophrenia

JAMA Psychiatry75749751

10.1001/jamapsychiatry.2018.0844

29799905

Kim

Racine

Adluru

Hwang

Blennow

Zetterberg

Carlsson

Asthana

Koscik

Johnson

Bendlin

Singh

2019

Cerebrospinal fluid biomarkers of neurofibrillary tangles and synaptic dysfunction are associated with longitudinal decline in white matter connectivity: a multi-resolution graph analysis

NeuroImage: Clinical21

101586

10.1016/j.nicl.2018.10.024

30502079

La Joie

Visani

Baker

Brown

Bourakova

Cha

Chaudhary

Edwards

Iaccarino

Janabi

Lesman-Segev

Miller

Perry

O'Neil

Pham

Rojas

Rosen

Seeley

Tsai

Miller

Jagust

Rabinovici

2020

Prospective longitudinal atrophy in Alzheimer's disease correlates with the intensity and topography of baseline tau-PET

Science Translational Medicine12

eaau5732

10.1126/scitranslmed.aau5732

31894103

Lin

Ren

Wang

Anthony

Tadin

Heffner

2017

Cortical thickness is associated with altered autonomic function in cognitively impaired and non-impaired older adults

The Journal of Physiology59569696978

10.1113/JP274714

28952161

Maass

Berron

Libby

Ranganath

Düzel

2015

Functional subregions of the human entorhinal cortex

eLife4

e06426

10.7554/eLife.06426

Maass

Berron

Harrison

Adams

La Joie

Baker

Mellinger

Bell

Swinnerton

Inglis

Rabinovici

Düzel

Jagust

2019

Alzheimer's pathology targets distinct memory networks in the ageing brain

Brain14224922509

10.1093/brain/awz154

31199481

Mitchell

Mufson

Schneider

Cochran

Nissanov

Han

Bienias

Lee

Trojanowski

Bennett

Arnold

2002

Parahippocampal tau pathology in healthy aging, mild cognitive impairment, and early alzheimer's disease

Annals of Neurology51182189

10.1002/ana.10086

11835374

Müller

Greverus

Weibrich

Dellani

Scheurich

Stoeter

Fellgiebel

2007

Diagnostic utility of hippocampal size and mean diffusivity in Amnestic MCI

Neurobiology of Aging28398403

10.1016/j.neurobiolaging.2006.01.009

16529847

Musiek

Holtzman

2015

Three dimensions of the amyloid hypothesis: time, space and 'wingmen'

Nature Neuroscience18800806

10.1038/nn.4018

26007213

Navarro Schröder

Haak

Zaragoza Jimenez

Beckmann

Doeller

2015

Functional topography of the human entorhinal cortex

eLife4

e06738

10.7554/eLife.06738

Ossenkoppele

Iaccarino

Schonhaut

Brown

La Joie

O'Neil

Janabi

Baker

Kramer

Gorno-Tempini

Miller

Rosen

Seeley

Jagust

Rabinovici

2019

Tau covariance patterns in Alzheimer's disease patients match intrinsic connectivity networks in the healthy brain

NeuroImage: Clinical23

101848

10.1016/j.nicl.2019.101848

31077982

Ousdal

Kaufmann

Kolskår

Vik

Wehling

Lundervold

Westlye

2020

Longitudinal stability of the brain functional connectome is associated with episodic memory performance in aging

Human Brain Mapping41697709

10.1002/hbm.24833

31652017

Ranganath

Ritchey

2012

Two cortical systems for memory-guided behaviour

Nature Reviews Neuroscience13713726

10.1038/nrn3338

22992647

Rose

Chen

Chalk

Zelaya

Strugnell

2000

Loss of connectivity in Alzheimer's disease: an evaluation of white matter tract integrity with colour coded MR diffusion tensor imaging

Journal of Neurology69528530

10.1136/jnnp.69.4.528

Salami

Wåhlin

Kaboodvand

Lundquist

Nyberg

2016

Longitudinal evidence for dissociation of anterior and posterior MTL Resting-State connectivity in aging: links to perfusion and memory

Cerebral Cortex2639533963

10.1093/cercor/bhw233

27522073

Schöll

Lockhart

Schonhaut

O'Neil

Janabi

Ossenkoppele

Baker

Vogel

Faria

Schwimmer

Rabinovici

Jagust

2016

PET imaging of tau deposition in the aging human brain

Neuron89971982

10.1016/j.neuron.2016.01.028

26938442

Smith

2002

Fast robust automated brain extraction

Human Brain Mapping17143155

10.1002/hbm.10062

12391568

Staffaroni

Brown

Casaletto

Elahi

Deng

Neuhaus

Cobigo

Mumford

Walters

Saloner

Karydas

Coppola

Rosen

Miller

Seeley

Kramer

2018

The longitudinal trajectory of default mode network connectivity in healthy older adults varies as a function of age and is associated with changes in episodic memory and processing speed

The Journal of Neuroscience3828092817

10.1523/JNEUROSCI.3067-17.2018

29440553

Suárez-Calvet

Morenas-Rodríguez

Kleinberger

Schlepckow

Araque Caballero

MÁ

Franzmeier

Capell

Fellerer

Nuscher

Eren

Levin

Deming

Piccio

Karch

Cruchaga

Shaw

Trojanowski

Weiner

Ewers

Haass

Alzheimer’s Disease Neuroimaging Initiative

2019

Early increase of CSF sTREM2 in Alzheimer's disease is associated with tau related-neurodegeneration but not with amyloid-β pathology

Molecular Neurodegeneration14

10.1186/s13024-018-0301-5

30630532

Wang

Men

Gao

Caramazza

2020

Two forms of knowledge representations in the human brain

Neuron107383393

10.1016/j.neuron.2020.04.010

32386524

Wang

Wedeen

2007

TrackVis. Org

Martinos Center for Biomedical Imaging0.6.1

Massachusetts

http://trackvis.org/

Xia

Wang

2013

BrainNet viewer: a network visualization tool for human brain connectomics

PLOS ONE8

e68910

10.1371/journal.pone.0068910

23861951

Lam

CLM

Lee

TMC

2017

White matter microstructural abnormalities in Amnestic mild cognitive impairment: a meta-analysis of whole-brain and ROI-based studies

Neuroscience & Biobehavioral Reviews83405416

10.1016/j.neubiorev.2017.10.026

29092777

Zhang

Descoteaux

Zhang

Girard

Chamberland

Dunson

Srivastava

Zhu

2018

Mapping population-based structural connectomes

NeuroImage172130145

10.1016/j.neuroimage.2017.12.064

29355769

10.7554/eLife.62114.sa1

Decision letter

Irish

Muireann

Reviewing EditorUniversity of SydneyAustralia

Kassady

Kaitlin

ReviewerUnited States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Chen and colleagues provide timely new insights into the link between CSF measures of phosphorylated tau, changes in brain structural connectivity over time, and cognitive decline in a sample of older adults who are cognitively normal or have mild cognitive impairment. The concept of "stability" used here provides a novel means by which changes in structural integrity can be assessed, and opens up new avenues to understand how tau impacts structural brain networks implicated in episodic memory decline in aging and Alzheimer's disease.

Decision letter after peer review:

Thank you for submitting your article "Longitudinal stability of medial temporal lobe connectivity explains tau-related memory decline" for consideration by eLife. Your article has been reviewed by Chris Baker as the Senior Editor, a Reviewing Editor, and three reviewers. The following individual involved in review of your submission has agreed to reveal their identity: Kaitlin Kassady (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data.

Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Summary:

Chen et al., present a timely study exploring the link between CSF measures of Ptau, changes in brain structural connectivity across 2 years (i.e., "stability"), and cognitive decline in a sample of older adults who are either cognitively normal or have mild cognitive impairment. While the earliest changes in structural connectivity were found within the MTL, and between the MTL and AT system, MTL and MTL-PM stability was related to better average episodic memory and change in episodic memory over five years. The research question of how tau impacts specific structural networks is important to further understand episodic memory impairment in aging and Alzheimer's disease, and the concept of "stability" provides a novel means by which to assess changes in structural integrity.

Essential revisions:

1) Introduction:

a) The authors state that hyperphosphorylated tau proteins are "necessary for subsequent Aβ aggregation". However, the prevailing view in the field is that Aβ aggregation is not dependent on concurrent tau pathology (ex. autosomal dominant AD). Please amend.

b) The "PM" system is referred to as "posterior-MTL", but there are no MTL regions in their PM network. It would be helpful to use the terminology "posterior-medial" when referring to "PM", as well as citing Ranganath and Ritchey, 2012, which first introduced this framework.

d) In the Introduction the authors state that the "MTL is more strongly connected to the AT than PM systems", however this is not accurate. In the papers they cite (especially Navarro Schröder, 2015), some regions of the MTL (such as parahippocampal gyrus and posteromedial entorhinal cortex) are more strongly connected to the PM system.

e) I believe Berron et al., showed mostly reduced MTL-AT functional connectivity in CU Ab+ vs CU Ab- and while there are some reduced connections between the MTL-AT with increased p-tau levels, increasing p-tau levels are much more robustly associated with reduced MTL-PM connectivity in CU Ab+.

f) Maass et al., report findings on functional task-based activity rather than functional connectivity

g) In the Introduction, the findings from Schröder et al., and Maass et al., are interpreted as indicating stronger associations between the MTL and the AT system. I believe these two papers report functional subregions in the entorhinal cortex that are differentially connected with two parallel processing streams (in the MTL and the neocortex) and the hippocampus, which means that the MTL is strongly connected with both the AT and PM systems.

h) The hypotheses in the Introduction could be formulated more clearly. What for example is meant by "tau pathology worsens"? It might be beneficial for the reader to refer to the stereotypical spatial progression pattern of tau accumulation.

i) In the fourth paragraph of the Introduction it should read, "An emerging study suggests that a functional connectome stability measure is associated with cognition (average and rate of episodic memory decline)."

Materials and methods and Results:

a) The rationale for the specific ROIs for MTL, AT, and PM networks is not clear, particularly as some of their selections differ from previous work. For example, the parahippocampal gyrus is a key region of the PM network, yet is placed in the MTL network here. Additionally, including more non-MTL AT regions, such as inferior temporal gyrus, could be beneficial to make the number of AT and PM regions more equivalent (3 regions/36 connections vs. 6 regions/60 connections), as more regions may inherently make the PM network more stable. Please cite appropriate references.

b) Could the authors include a visualization of the streamlines/connections between bilateral regions in order to understand whether that overlaps mainly with the aforementioned tracts. It would further help if the authors could elaborate in more detail on the underlying structural tracts, where they believe the signal is coming from and show examples of streamlines between MTL regions.

c) I was curious why the authors did not look at a pure baseline measure of memory performance at the first time point as a cleaner cross-sectional measure. Could the authors elaborate on that?

d) In the GLM, the authors report that their stability measure was associated with age (Model 2). However, they do not include age as a covariate in all other models. The same is true for other important covariates such as sex and years of education given that they are potentially related to CSF p-Tau and memory performance.

e) The authors should assess the relationships in models 3-5 while considering the effects of sex, age, and education. For model 5, given that stability is associated with age, how can the authors be sure that the moderating effect is driven by stability and not (also) by age? Regarding model 5, could the authors perform a mediation analysis to specifically test whether there is a significant indirect effect of CSF p-Tau via the stability measure on memory performance and decline while accounting for continuous age, sex, education and amyloid? In this framework the authors could also examine to what extent stability explains the effect of tau on memory (partial vs. full mediation).

f) "Stability" is interpreted as a positive aspect of structural integrity. However, it is possible that subjects with low MD measures at baseline stayed low at follow-up, and thus were "stable" (highly correlated) but have weak connectivity and low white matter integrity? This could be particularly common in the MTL in the MCI group, as damage from Ptau may have already occurred. Similarly, were there any subjects who had higher MD at follow-up, who were then "unstable" but increasing in connectivity (biologically less likely)? A breakdown of these different groups (high-MD stable, low-MD stable, high-low MD instable, low-high MD instable) and their prevalence in the CN/MCI samples would be informative.

g) Regarding the relationship between MTL structural stability, Ptau, and memory over 5 years in the whole sample – did you also test whether MTL-AT predicted average episodic memory? According to Figure 4, it appears that you did. Please include this info in the main text.

h) It would be helpful to have a figure (perhaps in the supplement) that shows the average mean diffusivity (MD) matrices for each group at each time point. This would allow readers to better understand the level of variability and raw values of these connectivity measures.

Discussion:

a) The finding of a stronger relationship between more Ptau and faster rate of memory decline in those with higher MTL stability (Figure 5) seems to run counter to the other findings where "stability" reflects structural integrity. Could this be explained by the "high stability" group containing both low-low MD and high-high MD subjects, allowing a strong relationship with cognition to emerge? I would like to see some consideration of this point in the Discussion.

b) In the stability group comparisons results (Figure 2A), the contrast between CN Ptau+ and MCI Ptau+ is currently not performed. From the figure, it appears that there may be a significant difference between those groups only in MTL-PM. This missing result would support their other findings because it would suggest that lower stability in MTL-PM connections might be related to the expression of clinical symptoms. Please amend and discuss accordingly.

c) While there was no interaction between CSF Aβ or neurodegeneration with Ptau on stability, was there a direct correlation between continuous levels of CSF Aβ or neurodegeneration with stability? If not, it helps support the finding that stability is specifically related to Ptau. Further, how do you interpret the age group interaction between Ptau and stability?

d) It is not clear why MTL-AT stability would not relate to episodic memory (average and rate of change) but MTL and MTL-PM would? Please discuss.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Longitudinal stability of medial temporal lobe connectivity explains tau-related memory decline" for further consideration by eLife. Your revised article has been evaluated by Chris Baker as the Senior Editor and a Reviewing Editor.

The manuscript has been much improved but there is one remaining issue that needs to be addressed before acceptance, as outlined below:

The reviewers are very pleased with the manner in which you have addressed all of their concerns, however, they believe the Title warrants modification prior to publication. There was concern that the term "explains" is perhaps too strong given that there are many other factors that were not taken into account that might explain the effect demonstrated here. The reviewers have therefore suggested substituting "moderates" or "is associated with" rather than "explains".

Similarly, please ensure that any causal or overly strong claims are tempered in the Abstract and Discussion.

10.7554/eLife.62114.sa2

Author response

Essential revisions:

1) Introduction:

Thanks for pointing this out. We further clarified in the Introduction: “The dominant view in the literature is that Aβ pathology triggers tau aggregation, strongly supported by evidence from autosomal dominant AD (Hardy and Selkoe, 2002). However, there have been important recent findings from sporadic AD that suggest that hyperphosphorylated tau proteins that are no longer governed by normal cellular removal mechanisms play a necessary and underappreciated role in AD progression (Musiek and Holtzman, 2015), with some authors even suggesting this may be the primary step in the disease needed for subsequent AB aggregation (Arnsten et al., 2020, Braak and Del Tredici, 2015).”

The reviewer is completely correct. We have now used "posterior-medial" when referring to “PM” and cited Ranganath and Ritchey, 2012 in the Introduction.

The reviewer is correct. We further clarified in the introduction: “MTL is strongly connected with both the AT and PM systems, with anterolateral entorhinal cortex showing preferential connectivity with anterior regions overlapping with the AT system and parahippocampal gyrus and posteromedial entorhinal cortex with posterior regions overlapping with the PM system (Maass et al., 2015, Schröder et al., 2015)”.

The reviewer is correct. To avoid confusion, we have taken this out from the Introduction.

f) Maass et al., report findings on functional task-based activity rather than functional connectivity.

We totally agree with review. We now modified the sentence in the Introduction:

“MTL is strongly connected with both the AT and PM systems, with anterolateral entorhinal cortex showing preferential connectivity with anterior regions overlapping with the AT system and parahippocampal gyrus and posteromedial entorhinal cortex with posterior regions overlapping with the PM system (Maass et al., 2015, Schröder et al., 2015).”

“Furthermore, tau pathology is related to worse performance in object memory tasks involving AT system relative to scene memory tasks involving PM system in cognitive normal older adults, matching the proposed spread of the disease via neuronal connections, as tau burden was much higher in the AT system compared with PM system in the early stages of AD (Arnsten et al., 2020, Maass et al., 2019).”

We further clarified in the Introduction:

“Second, given that tau tangles deposit early in the anterior temporal lobe (Braak and Del Tredici 2015, Maass et al., 2019), we hypothesized that Ptau first affects stability in the MTL and MTL-AT in the predementia stages, followed by MTL-PM as the disease progresses and tau pathology worsens and spreads to posterior brain regions”.

Thanks for pointing this out. We have corrected it.

Materials and methods and Results:

This is a valid concern. Numerous studies have shown that medial temporal lobe (MTL) structures including the hippocampus and the surrounding hippocampal region consisting of the parahippocampal and entorhinal neocortical regions, are crucial for long-term memory. In terms of brain anatomy, parahippocampal gyrus receives input mainly from visuospatial association areas and provides major input to the entorhinal cortex, which in turn feeds directly into the hippocampus (Davachi et al., 2003). At the functional level, parahippocampal gyrus plays an important role in episodic memory relating to associative memory, source memory, and recollection (Aminoff et al., 2013). Given the anatomy and function of the parahippocampal gyrus, we consider it to belong to the MTL memory network.

We further clarified the rationale in the Results: “Since the hippocampus and the surrounding hippocampal region including the parahippocampal cortex and entorhinal cortex, are the primary regions supporting memory processing (Davachi et al., 2003), we consider them to belong to the MTL memory network. Thus, the MTL system includes entorhinal, hippocampus, and parahippocampal gyrus in both hemispheres.”

We included the bilateral inferior temporal gyrus into the non-MTL AT regions as the reviewer suggested and updated the results. We now added in the method: “In line with previous literature (Berron et a.,l 2020, Ranganath and Ritchey 2012), the AT system includes bilateral inferior temporal, temporal polar, lateral and medial orbitofrontal cortex, while the PM system includes bilateral posterior and isthmus cingulate, lateral occipital cortex, precuneus, and thalamus.”

Thanks for pointing this out. We have added the Figure 1B to visualize the streamlines between MTL related regions in a representative participant. We now added in the Results:

“For visualization purposes, we presented the connections generated between MTL related structures in a representative participant in Figure 1B. The connections within MTL include the hippocampal cingulum bundle and fornix. The connections between MTL and AT largely overlap with the anterior segments of the inferior longitudinal fasciculus (ILF) and inferior fronto-occipital fasciculus (IFOF), while the connections between MTL and PM mainly involve the cingulum bundle and posterior segments of ILF and IFOF.”

c) I was curious why the authors did not look at a pure baseline measure of memory performance at the first time point as a cleaner cross-sectional measure. Could the authors elaborate on that?

The stability measure is based on the information from both baseline and 2 year follow up. It’s more clinically meaningful to predict cognition in the future (up to four year follow-up) rather than the baseline collected prior to the stability measure. We now added in the Materials and methods:

“To assess the association between the longitudinal structural stability, baseline Ptau, and episodic memory, an average episodic memory score is calculated by averaging episodic memory over 5 years. Since we were interested in prediction of memory in the future, average memory score was used instead of baseline memory”.

Thanks for pointing this out. We have now examined relationships between the longitudinal structural stability, baseline Ptau, and episodic memory in the whole sample, including and excluding the covariates (age, education, sex, CSF Aβ, and neurodegeneration). Changes have been highlighted in yellow in the method and result sections.

For model 5 (referred as model 6 in the revised manuscript), we now show that the moderating effect of MTL stability remained after the covariates were controlled (age, education, sex, CSF Aβ, and neurodegeneration).

We conducted the mediation analysis and didn’t find a significant indirect effect of CSF p-Tau via the stability measure on memory, or rate of memory change. Regardless of the lack of mediation effect, we did find a significant moderation effect of MTL stability on the relationship between tau and memory decline rate (see Model 6’ and Figure 6). To keep the paper more focused, we didn’t report these negative findings.

First, we want to clarify that the stability is a similarity measure, which reflects information in the patterns of MD across multiple connections, rather than only changes in mean MD. We agree with the reviewer that it’s possible that some subjects with low average MD at baseline and follow-up could show relatively high stability, while other could show low stability if their MD patterns changed across specific connections. We also want to clarify that, while it is possible that a loss of integrity causes poor stability due to an overall reduction in connections caused by pathology, it is also possible that pathology in certain connections could cause them to decrease connectivity, while other connections show corresponding increases as a result. Stability as a similarity measure is ambiguous to the mechanism by which pathology causes connections to change, but we believe that change in the overall pattern of connectivity in itself is meaningful.

To address the reviewer’s concern, we averaged the MD measures across all connections within MTL to get a mean MD score for MTL network. Then we used the median as a cutoff to dichotomize the mean MD score to “high” and “low”. Similarly, we dichotomized stability as “stable” vs. “instable”. The Table 2 in the manuscript displays the prevalence in the CN/MCI samples.

As we can see from Table 2, the prevalence of low-low stable individuals is actually not common in the MCI group (13.3%). In the Materials and methods, we now added:

“The traditional univariate analysis only considers the overall magnitude of the network (e.g., the average MD across all connections within MTL network). […] Thus, the traditional univariate analysis may not be as sensitive as the stability measure in detecting these subtle structural changes in the predementia AD stage.”

Thanks for pointing this out. We did test whether MTL-AT predicted average episodic memory and we have now included this result in the main text and Figure 5 middle.

This is a good point. We have now added this figure in Figure 1—figure supplement 1.

Discussion:

This is a valid concern. As can be seen in Table 2, the stable group contains 12 high-high MD and 15 low-low MD subjects, while the instable group contains 12 high-high MD and 9 low-low MD subjects. As described above, since stability is calculated using all connections, rather than mean MD, instability is not restricted to high-low or low-high MD subjects. We now added in the Discussion:

“It is possible that the observation of a stronger relationship between more Ptau and faster rate of memory decline in those with higher MTL stability might be driven by greater variance in this group, caused by the inclusion of both stable individuals with overall low connectivity as well as consistently high connectivity. To assess this, we looked at the number of spread of individuals across these groups based on their mean structural connectivity (Table 2). The stable group contains 12 high-high MD and 15 low-low MD subjects, while the instable group contains 12 high-high MD and 9 low-low MD subjects. While there are more low-low subjects in the stable group, the ratio of low-low to high-high is similar across these groups, suggesting increased variance is unlikely to drive the strong relationship with cognition in the high stability group.”

This is an excellent point. We have conducted the contrast between CN Ptau+ and MCI Ptau+ and updated the results and discussion accordingly.

In the Results, we modified the sentences: “In line with our hypothesis that MTL and MTL-AT stability is affected by Ptau early in the predementia AD stage, compared to CN Ptau-, structural stability was significantly lower in CN Ptau+ in MTL (t(42)=-2.11, p = 0.022) and MTL-AT (t(42)=-2.57, p = 0.007). Decreased structural stability was further found in MTL (t(54)=-2.61, p = 0.006) and MTL-AT (t(54)=-3.55, p < 0.001) in MCI Ptau+, compared to CN Ptau-. In line with our hypothesis that MTL-PM is affected later in disease progression, we found a significantly lower stability in MTL-PM in MCI Ptau+, compared to CN Ptau+ (t(46)=-5.12, p < 0.001) and CN Ptau- (t(54)=-3.89, p < 0.001, Figure 2A).”

In Discussion, we modified the sentence: “CN Ptau+ and MCI Ptau+ had decreased stability in MTL and MTL-AT compared to CN Ptau-, while MCI Ptau+ further showed decrease in MTL-PM stability relative to both CN Ptau- and CN Ptau-+ groups, suggesting that Ptau relates to structural stability in a temporal pattern that matches its proposed spread from the MTL via neuronal connections”.

We have now examined the direct correlation between continuous levels of age, CSF Aβ or neurodegeneration with stability in model 2 and updated the result in the main text and Figure 3. In the Results, we added now:

“We examined whether structural stability could be directly affected by covariates (i.e., age, CSF Aβ or neurodegeneration) in the entire sample with Model 2 (Y stability = β20 + β21Covariates + ԑ₂). We found no significant relationship between age and stability (all ps > 0.05, Figure 3A). We found a significant correlation between CSF Aβ and stability in MTL-PM (β = 0.40, SE = 0.16, Wald’s χ2 = 6.04, BH-adjusted p = 0.042, Figure 3B). Neurodegeneration significantly related to stability within MTL (β = 0.58, SE = 0.26, Wald’s χ² = 4.97, BH-adjusted p = 0.039), MTL-AT (β = 0.49, SE = 0.25, Wald’s χ² = 3.89, BH-adjusted p = 0.048) and MTL-PM (β = 0.61, SE = 0.19, Wald’s χ² = 10.39, BH-adjusted p = 0.003, Figure 3C).”

Regarding the age group interaction between Ptau and stability, after we included the inferior temporal cortex into AT regions, there was no significant age group interaction (Figure 4A).

d) It is not clear why MTL-AT stability would not relate to episodic memory (average and rate of change) but MTL and MTL-PM would? Please discuss.

After we included the inferior temporal cortex into the non-MTL AT regions as one reviewer suggested, MTL-AT stability predicted episodic memory (average episodic memory). We have updated the Results in the main text and Figure 5.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been much improved but there is one remaining issue that needs to be addressed before acceptance, as outlined below:

We totally agree. We now used "is associated with" instead of “explains”. The current title is “Longitudinal stability of medial temporal lobe connectivity is associated with tau-related memory decline”.

Similarly, please ensure that any causal or overly strong claims are tempered in the Abstract and Discussion.

We have altered the claims and made them less causal in the abstract, introduction and discussion:

In the Abstract, we wrote: “…this loss of stability in connections known to be important for memory moderated the relationship between Ptau accumulation and memory decline.”

In the Introduction, we wrote: “Finally, given that MTL and related structures play a central role in episodic memory, we hypothesized that MTL related structural stability might be associated with tau-related episodic memory decline.”

In the Discussion, we wrote: “… whether this loss of stability in connections important for episodic memory is related to the link between Ptau accumulation and cognitive decline.”

“Importantly, structural stability in MTL moderated the effect of Ptau on the rate of memory change, suggesting that structural stability in MTL may be a candidate as a mechanistic link between the accumulation of Ptau and cognitive decline.”

“A model showed that decreases in the stability of within MTL connections significantly predicted episodic memory decline, suggesting changes in these connections resulting from increased Ptau is associated with decreases in episodic memory.”