Vulnerability of the locus coeruleus–entorhinal cortex white matter tract in autosomal dominant Alzheimer's disease

Abstract

INTRODUCTION

Locus coeruleus (LC) structural integrity declines early in sporadic and autosomal dominant Alzheimer's disease (AD). We examined LC–entorhinal cortex (LC‐EC) tract integrity and its association with amyloid and tau pathology and memory in autosomal dominant AD (ADAD).

METHODS

We examined associations between diffusion‐weighted imaging‐derived LC‐EC tract characteristics and age, sex, amyloid (Pittsburgh Compound B [PiB]), entorhinal tau ([¹⁸F]‐flortaucipir [¹⁸F‐FTP] positron emission tomography [PET]), and memory in PSEN1‐E280A mutation carriers (n = 30) and non‐carriers (n = 33).

RESULTS

LC‐EC tract integrity was lower in females and older individuals in both non‐carriers and carriers. In mutation carriers, lower tract integrity was associated with higher entorhinal tau and poorer memory, with tau and amyloid mediating the tract‐memory relationship in the entire sample but not in unimpaired carriers.

DISCUSSION

These findings highlight the LC‐EC tract as an early disease vulnerability marker in ADAD, with tract disruption associated with memory decline across disease stages, a relationship further mediated by entorhinal tau pathology as disease progresses.

Highlights

• Locus coeruleus–entorhinal cortex (LC‐EC) tract integrity is a structural marker of disease progression in autosomal dominant Alzheimer's disease (ADAD).

• Women show lower LC‐EC tract integrity than men, independent of mutation status.

• LC‐EC tract integrity in carriers is negatively associated with entorhinal tau positron emission tomography (PET).

• Tau and amyloid PET mediate the tract‐memory relationship in mutation carriers.

• LC‐EC tract integrity is associated with memory in cognitively unimpaired carriers

1. BACKGROUND

Alzheimer's disease (AD) pathology is characterized by the accumulation of beta‐amyloid (Aβ) plaques and neurofibrillary tau tangles in the brain, starting decades before clinical symptoms emerge. ¹ , ² While the majority of AD cases are sporadic, a subset of individuals are carriers of deterministic genetic mutations causing early‐onset AD. In carriers of the autosomal dominant presenilin1 (PSEN‐1) E280A mutation, first identified in a large Colombian kindred, ³ mild cognitive impairment typically emerges around age 44, and individuals progress to dementia by age 50. ⁴ However, memory changes are evident over a decade before clinical disease onset in carriers of autosomal dominant AD (ADAD). ³

The PSEN‐1 E280A mutation leads to increased production of Aβ‐42 and has been associated with higher levels of hyperphosphorylated tau and neurofibrillary tangle (NFT) formation compared to sporadic forms of AD. ⁵ , ⁶ , ⁷ , ⁸ , ⁹ Autopsy studies in the general population reveal that the locus coeruleus (LC) is one of the earliest brain regions with detectable tau pathology in AD, ² , ¹⁰ , ¹¹ , ¹² , ¹³ before tau becomes detectable in the entorhinal cortex (EC), the first cortical region with tau pathology. ² Tau pathology is present in LC neurons from Braak stage 0, ¹² which is accompanied by progressive volumetric loss, ¹⁴ , ¹⁵ including substantial neuronal loss from Braak stage III onward. ¹³ , ¹⁴ , ¹⁶ Post mortem and in‐vivo magnetic resonance imaging (MRI) studies of PSEN‐1 mutation carriers reported similar observations, with neuronal loss, tau pathology, and structural changes in the LC as a consistent feature of ADAD. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰

The early vulnerability of the LC is particularly consequential because it serves as the brain's primary source of norepinephrine (NE), a neurotransmitter critical for attention and memory, cognitive domains affected in AD. In‐vivo neuroimaging studies of the LC in ADAD have shown that MRI‐derived LC integrity wanes as individuals approach the estimated age of disease onset, and lower LC integrity correlates with increasing cortical tau and worse cognitive performance. ¹⁹ , ²⁰

While the mechanisms underlying the tau progression from the LC to the EC remain under investigation, the projections forming the LC‐EC tract have been hypothesized as an early pathway for tau propagation in AD. ¹² , ²¹ , ²² These projections are vulnerable to early tau pathology, with phosphorylated tau present in LC axons as early as Braak stage 0, ¹² and an animal model of AD demonstrating that LC tau pathology is associated with degeneration of noradrenergic terminals in the EC. ²³ In‐vivo human diffusion MRI studies have further provided support for the vulnerability of the LC‐EC tract to AD pathology and neurodegeneration, reporting microstructural degeneration with tau pathology, cognitive impairment, and neurodegeneration markers. ²⁴ , ²⁵ , ²⁶ , ²⁷ Despite evidence from sporadic AD linking LC‐EC tract integrity to entorhinal tau pathology and cognitive decline, these relationships remain unexplored in individuals with ADAD, in whom the LC is affected very early on in the lifespan.

The clinical and pathological characteristics of ADAD closely resemble those of sporadic and late‐onset AD, with differences in the age of onset and temporal ordering of pathology. In PSEN‐1 E280A mutation carriers, PET‐based Aβ accumulation is detectable approximately 16 years before disease onset, ²⁸ , ²⁹ , ³⁰ while PET‐based tau pathology emerges 9 years prior to disease onset and correlates more strongly with cognitive decline, ²⁹ , ³⁰ , ³¹ providing a valuable window for studying early disease processes. As the disease trajectory in ADAD follows a more predictable pattern, this population provides a unique opportunity to investigate preclinical disease mechanisms without the confounding effects of aging and age‐related comorbidities (e.g., high blood pressure, diabetes ³² ).

In the current study, we aimed to characterize properties of the LC‐EC tract of PSEN‐1 E280A mutation carriers. We hypothesized that early entorhinal tau deposition is associated with disrupted LC‐EC tract integrity, which in turn contributes to lower memory performance. We employed fixel‐based analysis, which allowed us to accurately model fiber populations in voxels with complex and crossing fibers, and extract fiber‐specific properties of the LC‐EC tract using diffusion MRI data.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed the literature using traditional (e.g., PubMed) sources and meeting abstracts and presentations. Whereas the locus coeruleus (LC) to entorhinal cortex tract (EC) has not been studied in autosomal dominant Alzheimer's disease (ADAD), there have been several publications studying this tract in sporadic or late‐onset AD. The relevant citations are appropriately cited.

2. Interpretation: Our findings demonstrate that LC‐EC tract integrity is an early upstream structural marker associated with tau pathology and memory function in non‐demented ADAD individuals.

3. Future directions: These findings suggest that LC‐EC integrity may serve as a structural brain marker indicating vulnerability to disease progression in presymptomatic and early symptomatic disease stages. Future longitudinal studies examining LC‐EC tract changes in relation to disease trajectories will be critical to determine temporal ordering of events.

2. METHODS

2.1. Study population

In this retrospective study, we analyzed data from members of the Colombian kindred carrying the PSEN‐1 E280A mutation who were enrolled in the Colombia‐Boston (COLBOS) biomarker research study. The study sample included 63 individuals, of whom 30 were mutation carriers and 33 were non‐carrier relatives descended from the same ancestor. All participants underwent diffusion weighted imaging (DWI), [11C] Pittsburgh Compound B (PiB) positron emission tomography (PET), to measure in‐vivo Aβ burden, [18F]‐flortaucipir (FTP) PET, to quantify tau pathology, a T1‐weighted MRI scan to extract regions of interest, and cognitive testing. Owing to the highly predictable clinical course in this population, expected years to disease onset (EYO) were calculated for mutation carriers by subtracting each participant's age from the median age of clinical onset of disease (44 years of age). ⁴

2.2. Standard protocol approvals, registrations, and patient consent

The institutional review boards of the University of Antioquia, Colombia, and Mass General Brigham approved the study. All participants were 18 years of age or older at the time of enrollment and provided written informed consent.

2.3. Data processing

2.3.1. MRI protocol

MRI scans were conducted at Massachusetts General Hospital (MGH) on a 3T Siemens Trio scanner, using a 32‐channel head coil. Single shot spin echo planar imaging sequence DWI parameters: TR = 8040 ms, TE = 84 ms, FOV = 256 × 256 mm, 64 slices, and 2 × 2 × 2 mm voxels. A total of 30 non‐collinear gradient directions were applied with a b‐value of 700 s/mm², along with 5 non‐diffusion‐weighted images (b = 0 s/mm²). Accelerated T1‐weighted magnetization‐prepared rapid acquisition gradient echo and generalized autocalibrating partially parallel acquisition (MPRAGE GRAPPA) scans were obtained in the same session (repetition time [TR] = 2300 ms, echo time [TE] = 2.95 ms, 1.1 × 1.1 × 1.2 mm voxels) and processed using FreeSurfer 6 to extract region of interest parcellations. In a subset of individuals with diffusion imaging, LC specific MRI scans were acquired with a 2D‐T1‐weighted turbo‐spin‐echo (TSE) sequence with magnetization transfer (MT) contrast (TR = 743 ms, TE = 16 ms, six slices, 0.4×0.4×3.0 mm resolution). For a detailed description of the processing steps of the MT‐TSE scans and the extraction of the LC integrity values, see Jacobs et al., 2023. ²⁰

2.3.2. PET protocol

PET scans were acquired at MGH on a Siemens CTI ECAT PET HR+ scanner. PiB‐PET scans (69 frames, 60‐minute dynamic acquisition) measured in‐vivo amyloid burden. PiB‐PET images were converted to distribution volume ratio (DVR) referenced to cerebellar gray matter tracer uptake, using the Logan graphical method to extract region‐of‐interest specific DVRs using regional time‐activity curves. ³³ , ³⁴ FTP‐PET scans (80 – 100 minutes after bolus injection, 4 × 5‐minute frames) measured tau burden, expressed as standardized uptake value ratio (SUVR) referenced to the cerebellar gray matter derived from FreeSurfer segmentation. ³⁵ PET images were reconstructed, attenuation corrected and inspected for motion and adequate counts. Partial volume correction (PVC) was applied to the PET data using a geometric transfer matrix (GTM) method. A detailed description of the PET parameters can be found in Quiroz et al., 2018. ³⁰ MRI and PET scans were acquired within the same week.

2.3.3. DWI processing

Diffusion images were processed using MRtrix 3.0.4 for fixel‐based analysis. ³⁶ Fixels represent fiber bundle elements within a voxel consisting of fibers with similar orientation. Preprocessing steps included denoising, motion correction, FSL‐based distortion correction, bias field correction (ANTS N4), and global intensity normalization. Images were upsampled to 1.25 mm isotropic voxels. Fiber orientation distributions were estimated using constrained spherical deconvolution, ³⁷ , ³⁸ registered to a population template, ³⁹ and segmented to extract fixel metrics: fiber density (FD), fiber‐bundle cross‐section (logFC), and the combined metric (FDC). These metrics are informative on the information‐carrying capacity of white matter tissue with FD values reflecting the volume of restricted water in a voxel and logFC values reflecting changes in the cross‐sectional area occupied by the tract of interest.

2.3.4. Tractography and regions of interest

Tractography in Mrtrix was performed on the population‐level fiber orientation distribution (FOD) template, and all regions of interest were defined in template space. The MNI152 1 mm template (FSL) was registered to the FOD template using ANTs (antsRegistrationSyN), and the resulting transformation matrices were applied to the LC meta mask, ¹⁹ Julich Brain atlas derived EC, ⁴⁰ and the Sun probabilistic LC‐EC tract atlas ²⁷ masks (antsApplyTransforms) to bring all masks into FOD template space. In the interest of reproducibility, we have used publicly available Montreal Neurological Institute (MNI) ‐based templates as seed and target regions.

Probabilistic tractography generating 20,000 streamlines was performed in MRtrix (tckgen) to reconstruct the LC‐EC tract, using the LC meta mask as the seed region. The target region consisted of the EC probabilistic mask derived from the Julich Brain Atlas, threshold at 0.5 probability. Streamlines were prevented from crossing across hemispheres by a midline exclusion mask of 1 mm in width. Only streamlines terminating in the EC were retained (tckedit).

To facilitate comparison of our results with existing literature and verify the robustness of our results, as a sensitivity analysis, we additionally report the FBA metrics extracted from the binarized probabilistic LC‐EC tract atlas published by Sun et al.,  ²⁷ thresholded at 0.1 probability and dilated by one voxel (Table S1). Visual inspection of the registration of the Sun mask with our FOD template revealed a shift not corresponding to the underlying white matter and the mask was shifted +1 voxel along the x‐axis using FSL to align with the underlying white matter skeleton. Detailed documentation of all Mrtrix processing steps, including algorithmic details and implementation, is publicly available at https://mrtrix.readthedocs.io. The seed and target regions of interest and the reconstructed LC‐EC tract are illustrated in Figure 1A, the tractogram‐to‐fixel extraction in Figure 1B, and a schematic of fixel‐based metric reductions relative to a normal fiber bundle in Figure 1C.

FIGURE 1.

LC‐EC tract reconstruction and fixel extraction. (A) The LC meta mask used as seed region (blue), Julich EC target region (orange), and the LC‐EC tract generated using probabilistic tractography in MRtrix3 (cyan) are projected into MNI standard space. Inset: isolated view of the LC‐EC tract generated using probabilistic tractography and refined to retain streamlines starting in the LC and terminating in the EC target region. (B) Sagittal view of the reconstructed LC‐EC tract (left), zoomed view of the entorhinal end of the tractogram overlaid on fixel map (middle), and binarized tract mask overlaid on the fixel map (right). The tractogram was converted to a fixel mask and binarized to identify fixels belonging to LC‐EC tract (blue lines) versus excluded fixels (black lines), with mean FD, logFC, and FDC values extracted per participant within this fixel mask. (C) Schematic illustration of fixel‐based metric reductions relative to a normal fiber bundle. EC, entorhinal cortex; FD, fiber density; FDC, combined fiber density and cross‐section; LC, locus coeruleus; logFC, fiber‐bundle cross‐section; MNI, Montreal Neurological Institute.

2.4. Cognitive testing

Clinical testing was performed in Spanish at the University of Antioquia (Medellín, Colombia), including the Mini‐Mental State Examination (MMSE), Clinical Dementia Rating (CDR), and the Consortium to Establish a Registry for Alzheimer's disease battery (CERAD) adapted to this Colombian kindred. ⁴¹ To assess relationships with memory, we used delayed recall performance from the CERAD word list test. Participants also completed the Functional Assessment Staging Test (FAST), which was used to classify clinical status. Individuals with a FAST score of 2 or less were considered cognitively unimpaired individuals and met the criterion of having an MMSE score of at least 26 or higher. Cognitive testing was performed within 3 months of the scan acquisitions.

2.5. Statistical analysis

Analyses were performed using RStudio (R version 4.2.3). The threshold for statistical significance was set at two‐sided α= 0.05. We used linear regression models to test associations between LC‐EC tract characteristics and age and sex, PET‐based markers of Aβ (PiB) and tau pathology (FTP), and cognitive performance. Age was not included as a covariate in models that included mutation carriers, due to its tight coupling with the disease trajectory. We included sex as a covariate in the pathology models, with the addition of years of education in models investigating cognition. Multiple comparisons correction was applied using the Benjamini–Hochberg false discovery rate (FDR) procedure ⁴² (α = 0.05) separate for each group, and the Tukey procedure for the LC‐EC tract integrity comparisons across non‐carriers, unimpaired and impaired carriers.

To facilitate interpretation of reported sex differences, logFC values were additionally converted to percentage changes using the mathematical relationship: $Percentagechange=(e^{logFC}-1)\ast 100$, where e represents Euler's number (≈2.718). This conversion allows for intuitive interpretation of fiber bundle differences, as logFC values represent logarithmic ratios rather than direct percentage changes. ⁴³

To characterize potential non‐linear relationships between LC‐EC tract metrics and age/estimated years to disease onset in mutation carriers, we conducted post‐hoc piecewise linear regression including 95% confidence intervals (CIs) for the breakpoint estimation. Inflection points were determined using segmented regression analysis with the segmented package in R, fitting a piecewise linear model as a function of age with the initial breakpoint estimate set at the median age of the sample. Given our interest in the earliest detectable changes in LC‐EC tract integrity during the preclinical disease stage, we restricted the unimpaired mutation carrier group to participants below the average age of disease onset (≤ 44 years). Visual inspection confirmed that the two unimpaired participants exceeding this threshold appeared to influence the model fit. Significant slope differences before and after the breakpoint were assessed with the Davies test. Sex was included as a covariate given its significant association with LC‐EC logFC.

Structural equation modeling (lavaan package) tested whether entorhinal FTP‐PET signal mediates the relationship between the LC‐EC tract and memory performance. The analysis employed bootstrapped CIs (5000 samples) to test the significance of the detected effects. Sex and years of education were included as covariates. Separate analyses were conducted for the full mutation carrier group and the cognitively unimpaired mutation carrier subgroup.

3. RESULTS

3.1. Demographics

The study included 63 individuals: 33 controls and 30 PSEN1‐E280A mutation carriers. Among mutation carriers, 23 were cognitively unimpaired and seven were mildly impaired. Participants ranged in age from 27.8 to 49.3 years of age at the time of the PET scans (mean ± SD = 36.96 ± 5.5), and 57% were female. The mean educational level in this sample was 10.5 ± 4.2 (range 2–17). Compared with non‐carriers, mutation carriers were older, had fewer years of formal education, scored lower on the MMSE, and were more likely to have cognitive impairment (Table 1). Mutation carriers also had higher PiB DVR values and elevated entorhinal FTP SUVR, with 80% of carriers classified as PiB‐positive (Figure S1).

TABLE 1.

Characteristics of participants.

Parameter	Non‐carriers, N = 33 *	E280A carriers, N = 30 *	p‐value †
Age	35.30 (4.52)	38.92 (6.06)	0.010
Sex			0.9
Male	15 (45.45%)	12 (40.00%)
Female	18 (54.55%)	18 (60.00%)
Education (years)	11.64 (3.98)	9.33 (4.19)	0.029
MMSE	28.91 (0.91)	27.10 (3.02)	0.003
FAST			<0.001
1	32 (96.97%)	16 (53.33%)
2	1 (3.03%)	7 (23.33%)
3	0 (0.00%)	6 (20.00%)
4	0 (0.00%)	1 (3.33%)
WLL Recall	7.82 (1.26)	5.17 (3.02)	< 0.001
PiB status			<0.001
PIB‐	33 (100.00%)	6 (20.00%)
PIB+	0 (0.00%)	24 (80.00%)
PiB FLR (DVR; PVC)	1.12 (0.05)	1.71 (0.38)	<0.001
FTP entorhinal (SUVR; PVC)	1.06 (0.22)	1.74 (0.69)	<0.001

Note: Per PSEN1‐E280A mutation carrier status, the mean and standard deviation are given for age, education level, MMSE score, PiB, and entorhinal tau. The count and percentages are given for sex, PiB status, and the FAST which was used to define clinical status.

Abbreviations: DVR, distribution volume ratio; FAST, Functional Assessment Staging Test; FLR, frontal, lateral temporal, and retrosplenial; FTP, flortaucipir; MMSE, Mini‐Mental State Examination; PiB, Pittsburgh Compound B; PVC, partial volume correction; SUVR, standardized uptake value ratio;

^*Mean (SD); n (%);

^†One‐way analysis of means (not assuming equal variances); Pearson's chi‐squared test; Fisher's exact test.

3.2. LC‐EC tract associations with age and sex

LC‐EC tract fiber‐bundle cross‐section (logFC) was significantly lower in females than in males (Figure 2A), and within‐group analyses showed that this effect is independent of mutation carrier status (non‐carriers: t(31) = ‐4.47, p_FDR  = 0.0006; mutation carriers: t(28) = ‐4.38, p_FDR  = 0.0009; Table 2). On average, the fiber bundle cross‐sections were 8.7 % smaller in female non‐carriers, and 10.1% smaller in female carriers compared to males. Older individuals in the non‐carrier and carrier groups had lower logFC and FDC metrics (logFC: non‐carriers: t = ‐2.71, p_FDR  = 0.022; mutation carriers: t = ‐2.88, p_FDR  = 0.023; FDC: non‐carriers: t = ‐2.88, p_FDR  = 0.021; mutation carriers: t = ‐2.64, p_FDR  = 0.027), while FD alone did not reach significance (non‐carriers: t = ‐1.96, p_FDR  = 0.088; mutation carriers: t = ‐1.93, p_FDR  = 0.096; Table 2; Figure 2B–D).

FIGURE 2.

LC‐EC tract shows sex differences, and age‐related decline across groups. (A) LC‐EC tract logFC by sex, colored by group. Significant sex differences were observed in the entire sample and within each group (non‐carriers/carriers), with females showing lower logFC values than males. (B) In both carriers and non‐carriers, we observed a linear association between age (bottom x‐axis) and EYO (top x‐axis) and LC‐EC tract logFC, with lower LC‐EC tract integrity values in older individuals. (C) Linear association between LC‐EC tract FD metric and age (bottom x‐axis) and EYO (top x‐axis) did not reach significance in carriers and non‐carriers. (D) In both carriers and non‐carriers, we observed a linear association between age (bottom x‐axis) and estimated years to disease onset (EYO, top x‐axis) and LC‐EC tract FDC, with lower LC‐EC tract integrity values in older individuals. Lines show fitted linear regression model with 95% confidence intervals. EC, entorhinal cortex; EYO, estimated years to disease onset; FD, fiber density; FDC, combined fiber density and cross‐section; LC, locus coeruleus; logFC, fiber‐bundle cross‐section.

TABLE 2.

LC‐EC white matter tract structural metrics with age and sex split for mutation carriers and non‐carriers.

			Sex				Age
Group		df	β	t	puncor	pFDR	β	t	puncor	pFDR
Mutation carriers	FD	28	0.006	0.28	0.781	0.781	−0.003	−1.93	0.0643	0.096
	logFC	28	−0.090	−4.38	0.0002	0.0009*	−0.006	−2.88	0.0076	0.023*
	FDC	28	−0.034	−1.20	0.242	0.290	−0.006	−2.64	0.0135	0.027*
Non‐carriers	FD	31	0.010	0.50	0.622	0.622	−0.004	−1.96	0.0588	0.088
	logFC	31	−0.077	−4.47	0.0001	0.0006*	−0.006	−2.71	0.0108	0.022*
	FDC	31	−0.019	−0.72	0.479	0.575	−0.008	−2.88	0.0071	0.021*

Abbreviations: EC, entorhinal cortex; FD, fiber density; logFC, fiber bundle cross‐section; FDC, combined fiber density and cross‐section metric; LC, locus coeruleus.

Given the early involvement of the LC in AD pathophysiology, we examined the earliest detectable changes in LC‐EC tract integrity in unimpaired mutation carriers (n = 21). Segmented regression identified a potential breakpoint at age 31 (95% CI 28.8‐33.4) for the LC‐EC tract logFC metric, although this did not reach statistical significance (Davies test p = 0.079; Sup Figure S2), likely reflecting our limited sample size. The slope following this potential breakpoint indicated a negative age‐related trajectory in LC‐EC tract logFC (β = ‐0.009, t(17) = ‐3.07, p = 0.007), whereas the slope preceding the potential breakpoint is not significant (β = 0.031, t(17) = 1.60, p = 0.108). The breakpoint analyses for FD (p = 0.310) and FDC (p = 0.367) metrics were not significant.

3.3. LC‐EC tract associations with clinical and amyloid status

Group differences in LC‐EC tract logFC were significant for logFC (F(1,60) = 3.29, p = 0.044), but not for FD or FDC (Sup Figure S3). Pairwise comparisons showed that impaired carriers had significantly lower LC‐EC tract logFC than non‐carriers (logFC: β = 0.067, t(60) = 2.52, p_Tukey  = 0.037), while the difference between impaired and unimpaired carriers approached but did not reach significance (logFC: β = 0.064, t(60) = 2.31, p_Tukey  = 0.060). No significant pairwise differences were observed for FD or FDC.

In terms  of amyloid status, PiB‐positive carriers showed a negative association between age and all three LC‐EC tract metrics (FD: β = ‐0.004, t(22) = ‐2.16, p_FDR  = 0.042; logFC: β = ‐0.007, t(22) = ‐2.77, p_FDR  = 0.017; FDC: β = ‐0.008, t(22) = ‐3.05, p_FDR  = 0.017), whereas this association was not observed in PiB‐negative mutation carriers, although this group was small with n = 6 (FD: β = 0.009, t(4) = 1.09, p_FDR  = 0.337; logFC: β = 0.008, t(4) = 0.15, p_FDR  = 0.239; FDC: β = 0.015, t(4) = 0.16, p_FDR  = 0.239; Sup Figure S4). These findings suggest that LC‐EC tract logFC is associated with clinical impairment, and age‐related decline is most pronounced in mutation carriers with elevated amyloid burden.

3.4. LC‐EC tract integrity is associated with tau pathology in mutation carriers

Interaction analysis showed that lower LC‐EC tract fiber density and cross‐section metrics (FD; logFC; FDC) were associated with higher entorhinal FTP PET signal in mutation carriers but not in non‐carriers (FD: t(58) = ‐2.57, p_FDR  = 0.019; logFC: t(58) = ‐3.02, p_FDR  = 0.011; FDC: t(58) = ‐3.24, p_FDR  = 0.011; Sup Table S2; Figure 3A–C). Similarly, lower LC‐EC tract cross‐section and combined metrics (logFC; FDC), but not FD, were associated with higher PiB levels in mutation carriers relative to non‐carriers (FD: t(58) = ‐1.93, p_FDR  = 0.059; logFC: t(58) = ‐2.81, p_FDR  = 0.013; FDC: t(58) = ‐2.49, p_FDR  = 0.019; Figure 3D–F). Within mutation carriers, higher entorhinal FTP PET signal was associated with lower LC‐EC tract integrity across all three metrics, whereas associations with PiB levels did not survive FDR correction (Table 3). No significant associations were observed between LC‐EC tract metrics and either FTP or PiB in non‐mutation carriers.

FIGURE 3.

LC‐EC tract shows vulnerability to AD pathology. (A‐C) Interaction plots showing the relationship between entorhinal tau PET (FTP) and LC‐EC tract metrics by group (mutation carriers [blue triangles] vs. non‐carriers [gray circles]). (D‐F) Interaction plots showing the relationship between PiB FLR and LC‐EC tract metrics by group (mutation carriers [green triangles] vs. non‐carriers [gray circles]). Lines represent fitted linear regression models with 95% confidence intervals. Significant group by pathology interactions were observed for logFC and FDC. AD, Alzheimer's disease; EC, entorhinal cortex; FDC, combined fiber density and cross‐section; FLR, frontal, lateral temporal, and retrosplenial; FTP, flortaucipir; LC, locus coeruleus; logFC, fiber‐bundle cross‐section; PET, positron emission tomography; PiB, Pittsburgh Compound B.

TABLE 3.

LC‐EC white matter tract structural metrics with entorhinal tau‐PET (FTP) and amyloid‐PET (PiB) split for mutation carriers and non‐carriers, adjusted for the covariate sex.

			EC FTP				PiB FLR
Group		df	β	t	puncor	pFDR	β	t	puncor	pFDR
Mutation carriers	FD	27	−0.036	−2.62	0.0142	0.028*	−0.054	−2.01	0.0543	0.065
	logFC	27	−0.039	−2.77	0.0099	0.028*	−0.049	−1.78	0.0872	0.087
	FDC	27	−0.058	−3.03	0.0053	0.028*	−0.077	−2.06	0.0492	0.065
Non‐carriers	FD	30	0.010	0.22	0.831	0.992	−0.002	−0.01	0.992	0.992
	logFC	30	0.011	0.27	0.791	0.992	0.163	0.84	0.411	0.992
	FDC	30	0.010	0.151	0.881	0.992	0.088	0.29	0.773	0.992

Abbreviations: EC, entorhinal cortex; FD, fiber density; logFC, fiber bundle cross‐section; FDC, combined fiber density and cross‐section metric; FLR, frontal, lateral temporal, and retrosplenial; FTP, flortaucipir; LC, locus coeruleus; PET, positron emission tomography; PiB, Pittsburgh Compound B.

Restricting analyses to unimpaired mutation carriers resulted in similar associations with entorhinal FTP, although these did not survive FDR correction (FD: t(20) = ‐1.82, p_{unco r} = 0.083, p_FDR  = 0.083; logFC: t(20) = ‐2.45, p_{unco r} = 0.024, p_FDR  = 0.055; FDC: t(20) = ‐2.24, p_{unco r} = 0.037, p_FDR  = 0.055).

3.5. LC‐EC tract and memory performance

Interaction analyses showed significant group differences in the association between LC‐EC tract integrity and world list recall performance for the LC‐EC tract logFC (logFC: β = 25.23, t(57) = 3.70, p_FDR  = 0.001; Figure 4A) and the combined metric (FDC: β = 14.83, t(57) = 2.35, p_FDR  = 0.033; Figure 4B), but not for fiber density alone (FD: β = 11.91, t(57) = 1.30, p_FDR  = 0.198; Sup Table S3), indicating that logFC and FDC were associated with better word list recall performance in mutation carriers but not in non‐carriers. Within‐group analysis in mutation carriers confirmed the positive association between LC‐EC tract integrity and memory performance for logFC (logFC; β = 20.92, t(26) = 2.58, p_FDR  = 0.047; Sup Table S4), with this association also present in the subgroup of cognitively unimpaired mutation carriers (logFC: β = 17.85, t(19) = 3.29, p = 0.004).

FIGURE 4.

Entorhinal tau mediates the relationship between LC‐EC tract integrity and memory in the full mutation carrier group but not in unimpaired carriers. (A) Interaction plot showing the relationship between word list recall memory performance (WLL Recall) and LC‐EC tract logFC, and (B) combined FD and logFC (FDC) per group (mutation carriers (blue) vs. non‐carriers (gray)). (C) Mediation path diagram for FDC with standardized coefficients. (D) Bar graphs showing direct, indirect, and total effects from mediation models examining whether entorhinal tau (FTP‐derived) mediates the association between LC‐EC tract FDC and WLL Recall. Error bars represent 95% confidence intervals from bootstrap resampling (n = 5000). Significance levels: ^* p < 0.05, ^** p < 0.01. EC, entorhinal cortex; FD, fiber density; FDC, combined fiber density and cross‐section; FTP, flortaucipir; LC, locus coeruleus; logFC, fiber‐bundle cross‐section.

Since entorhinal FTP PET was associated with memory performance in mutation carriers (β = ‐2.70, t(26) = ‐4.82, p < 0.0001; covariates sex and education), we conducted mediation analyses and found that entorhinal FTP PET mediated the relationship between LC‐EC FDC and word list delayed recall performance in mutation carriers, accounting for 85% of the total effect (Figure 4C and D; Table 4). A significant mediation was also observed for logFC (56.7% mediation; Sup Table S5), whereas the total mediation model for FD did not reach significance (p = 0.142).

TABLE 4.

Mediation analysis of effect of LC‐EC tract combined fiber density and cross‐section on memory performance via entorhinal tau‐PET (FTP) in mutation carriers.

Path	Estimate	SE	CI lower	CI upper	p‐Value	Std estimate
c	2.177	5.239	−9.719	10.371	0.678	0.056
b	−2.595	0.756	−4.217	−1.178	0.001	−0.598
a	−4.778	1.342	−7.962	−2.643	<0.001	−0.535
Indirect	12.400	4.809	4.400	23.261	0.010	0.320
Total	14.577	5.468	2.769	24.000	0.008	0.376

Note: Results from structural equation modeling with bootstrapped confidence intervals (5000 samples), controlling for sex and education. N = 30. Percentage mediated = (indirect effect/total effect) × 100.

Abbreviations: CI, confidence interval; EC, entorhinal cortex; FTP, flortaucipir; LC, locus coeruleus; Path a, LC‐EC tract logFC → entorhinal tau; Path b, entorhinal tau → memory performance; Path c, direct effect of LC‐EC tract → memory performance; Indirect, mediated effect through entorhinal tau; PET, positron emission tomography; SE, standard error; Total, total effect of LC‐EC tract → memory performance.

Within the  unimpaired  mutation carriers, the total mediation model reached significance for logFC but not FDC (Sup Table S6‐7). Entorhinal FTP did not mediate the observed relationship between LC‐EC logFC and word list learning recall performance in unimpaired carriers (logFC: indirect effect β = 1.02, SE = 2.67, 95% CI [‐2.67, 8.12], p = 0.703). Whereas entorhinal FTP PET was associated with the LC‐EC logFC metric in unimpaired carriers, it was not associated with memory performance. However, the direct effect of LC‐EC tract logFC on memory performance remained significant after accounting for entorhinal FTP PET (Sup Table S6).

Repeating our mediation analysis using FLR amyloid PET (PiB) instead of entorhinal FTP PET resulted in similar findings with PiB mediating the relationship between LC‐EC tract FDC and logFC, and word list learning recall performance in the entire sample of mutation carriers (FDC: β = 9.17, SE = 3.23, 95% CI [3.60, 16.83], p = 0.005; percentage mediated: 63.9%; Sup Table S7; logFC: β = 10.66, SE = 4.02, 95% CI [3.45, 19.09], p = 0.008; percentage mediated: 44.5 %; Sup Table S8‐9), but within the unimpaired mutation carriers the total mediation model did not reach significance for either FDC or logFC (Sup Table S10‐11). The similar mediation patterns for PiB and FTP PET are consistent with their high intercorrelation in this population (r = 0.71, p < 0.0001).

3.6. LC‐EC tract integrity and LC integrity

We additionally examined whether LC‐EC tract metrics were associated with LC intensity values derived from MRI, available for a subset of our DWI sample (n = 31; n = 13 non‐carriers, n = 18 carriers). In mutation carriers, lower LC‐EC tract FD and the combined metric (FDC) were associated with lower LC intensity, reflecting degraded LC structural integrity (FD: β = 0.567, t(16) = 3.08, p_FDR  = 0.022; FDC: β = 0.345, t(16) = 2.59, p_FDR  = 0.030). The logFC metric did not reach significance (logFC: β = 0.243, t(16) = 1.04, p_FDR  = 0.314). The association between LC‐EC tract FD and LC intensity was also present in the subgroup of cognitively unimpaired mutation carriers (FD: β = 0.527, t(12) = 2.89, p = 0.014). No significant associations between LC intensity and the LC‐EC tract metrics were observed in non‐mutation carriers (Sup Table S12).

4. DISCUSSION

The  integrity of white matter pathways connecting brain regions vulnerable to early AD pathology may provide critical insights linking initial pathological changes to cognitive decline, as previous studies have suggested that tau pathology may spread through anatomical connections and disruption of these pathways may contribute to cognitive decline. In this cross‐sectional cohort study of individuals with ADAD, we focused on the LC‐EC tract integrity since the LC is among the earliest regions with tau pathology and its noradrenergic projections to the EC may represent a key pathway for early tau propagation and memory circuit disruption. Our findings show that lower LC‐EC tract integrity is associated with higher entorhinal tau pathology and worse memory performance, with both entorhinal tau and Aβ pathology mediating the tract‐memory relationships in mutation carriers overall but not in unimpaired mutation carriers. These findings suggest that lower LC‐EC tract integrity is linked to entorhinal tau early in the disease course, and the combined effects of tract disruption and tau pathology lead to memory impairment in symptomatic stages of ADAD. However, the association between lower LC‐EC tract integrity and worse memory performance in preclinical cognitively unimpaired mutation carriers suggests that disruption of this tract may impair the ability of the LC to regulate memory processes before clinical symptoms emerge.

Our  findings on LC‐EC tract integrity align with previous work suggesting that LC integrity may serve as a spatiotemporal predictor of cortical tau accumulation and cognitive decline, with declining LC integrity preceding cortical tau accrual in preclinical ADAD. ¹⁹ , ²⁰ Our results extend this work by demonstrating that structural integrity of the LC‐EC tract may contribute to cognitive decline in preclinical ADAD. Furthermore, lower LC integrity was associated with lower LC‐EC tract microstructure, with lower LC integrity corresponding to lower fiber density and combined fiber density and cross‐section. The observed association between LC integrity and LC‐EC tract fiber integrity is consistent with the observation that tau pathology affects LC axons early in the disease process and may reflect that tau‐related or other morphological changes in the LC result in downstream axonal loss along the LC‐EC pathway. Together, these findings position degeneration of the LC and LC‐EC tract as early contributors to cognitive changes in this cohort of ADAD, ⁴⁴ consistent with neuropathologic evidence that the LC and its projections are affected by tau and morphological changes early in the disease process, representing early upstream events in the AD cascade. ² , ¹¹ , ¹² , ⁴⁵ Additionally, the finding that females had smaller fiber bundle cross‐sections than males may inform differential disease vulnerability in AD risk and progression, warranting further investigation of sex‐specific changes in white matter tracts connecting brain regions vulnerable to early AD pathology.

In our ADAD cohort, lower LC‐EC tract integrity was associated with higher entorhinal tau burden and worse memory performance, with Aβ and entorhinal tau mediating the tract–memory relationship in the mutation carrier group, including individuals with cognitive decline. However, in cognitively unimpaired mutation carriers, the direct relationship between LC‐EC integrity and memory was not mediated by PET‐detectable Aβ or entorhinal tau pathology, despite significantly elevated PET levels compared to non‐carriers. These findings align with human neuroimaging findings in late onset AD. ²⁵ Given that tau pathology affects the LC early in the disease process, these disease‐stage dependent findings suggest that pathological progression and its interaction with LC‐EC tract integrity influence coupling to memory function. As cortical pathological burden increases in symptomatic individuals, the combined effects of accumulated tau and amyloid on both tract integrity and cognitive function become detectable as a mediation effect. In contrast, in preclinical cognitively unimpaired carriers, lower LC‐EC tract integrity may reflect early structural vulnerability or functional disruption affecting memory performance independent of PET‐detectable pathology burden, which is not yet driving memory performance in the absence of cognitive impairment. This aligns with findings in an animal AD model (TgF344_AD) where hyperphosphorylated tau in the LC correlated with progressive loss of noradrenergic innervations of the EC and memory impairment, demonstrating that early forms of tau in the LC affect the projections to the medial temporal lobe and memory function. ²³

Loss of LC‐noradrenergic innervation has profound consequences for memory processes, as NE enhances long‐term potentiation in the hippocampus ⁴⁶ , ⁴⁷ through adrenergic receptor activation, ⁴⁸ and facilitates synaptic strengthening necessary for memory consolidation. ⁴⁹ Beyond synaptic dysfunction, LC‐NE projection loss increases neuroinflammation through dysregulated microglial function. ⁵⁰ Furthermore, experimental and tau‐induced reductions in LC‐NE projections to the EC disrupt the medial perforant pathway. ⁵¹ , ⁵² the principal cortical input connecting the EC to the hippocampus, highlighting that reductions in LC‐EC projections can be associated with a cascade of pathological and cognitive changes in AD. Consistent with these findings, our in‐vivo diffusion imaging results showed that degeneration of the LC‐EC tract occurs early in the disease trajectory of ADAD and emerges in parallel with the earliest memory decline. ⁴⁴

Several  limitations should be considered. The cross‐sectional design limits causal inferences about the temporal relationships between LC‐EC changes, pathology, and cognition. Longitudinal studies are needed to confirm the proposed temporal sequence of pathology‐related LC‐EC tract changes. Diffusion data were acquired using a single‐shell protocol with a b‐value below 1000 s/mm², which may reduce the angular contrast of the diffusion signal and modestly attenuate the sensitivity of the fiber density metric. ⁵³ As this limitation applies to all our participants, this can potentially attenuate our findings making our observations more conservative estimates. The sample size was relatively small (n = 63), and to mitigate this, we implemented bootstrap resampling (n = 5000 iterations) to generate more robust CIs and assess the reliability of our findings. Our findings do not preclude the possibility that the presence of pretangle tau in LC neurons mediates early memory decline by disrupting axonal projections before cortical mature tangle pathology becomes PET‐detectable. Whereas the FTP signal in the EC reflects the aggregate burden of tangles, neuritic plaques, and neuropil threads, ⁵⁴ it does not give information on pretangle forms of tau. Finally, our findings are specific to PSEN1‐E280A carriers and may not generalize to other ADAD mutations or to sporadic forms of the disease, although recent work implicated this tract in relation to early tau pathology in preclinical sporadic AD. ⁵⁵

In conclusion, our findings demonstrate that LC‐EC tract integrity is associated with entorhinal tau pathology and memory performance in ADAD mutation carriers, with lower tract integrity contributing to memory decline across disease stages. This relationship is further mediated by PET‐detectable pathology burden when carriers with clinical symptoms are included, but not in unimpaired individuals suggesting that pathology links LC‐EC tract integrity and cognitive decline at more advanced disease stages. These findings highlight the functional importance of intact LC‐EC tract integrity for memory function and support the hypothesis that the LC‐EC tract is vulnerable to early tau pathology. Future longitudinal studies examining LC‐EC tract changes in relation to tau spreading and cognitive trajectories will be necessary to establish whether this pathway is a driver of early disease progression.

CONFLICT OF INTEREST STATEMENT

Yakeel T. Quiroz has served as consultant for Biogen. Heidi I.L Jacobs is the Immediate Past Chair of the ISTAART Neuromodulatory Subcortical Systems Professional Interest Area (PIA). Elouise A. Koops is the Programs Chair of the ISTAART Neuromodulatory Subcortical Systems PIA. David F. Aguillon has served on the advisory board of Eli Lilly Columbia. The other authors report no relevant conflicts of interest. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

All participants provided written informed consent before participating and were compensated. The study was approved by the University of Antioquia (Medellin, Colombia) and the Massachusetts General Hospital (Boston, USA) local institutional review boards.

Supporting information

Supporting Information

Supporting Information