Differential Vulnerability of Hippocampal Subfields to Amyloid and Tau Deposition in the Lewy Body Diseases

Abstract

Background and Objectives

Alzheimer disease (AD) copathologies of β-amyloid and tau are common in the Lewy body diseases (LBD), dementia with Lewy bodies (DLB) and Parkinson disease (PD), and target distinct hippocampal subfields compared with Lewy pathology, including subiculum and CA1. We investigated the hypothesis that AD copathologies impact the pattern of hippocampal subregion volume loss and cognitive function in LBD.

Methods

This was a cross-sectional and longitudinal, single-center, observational cohort study. Participants underwent neuropsychological testing and 3T-MRI with hippocampal segmentation using FreeSurferV7. PiB-PET and flortaucipir-PET imaging of comorbid β-amyloid (A) and tau (T) were acquired. The association of functional cognition, β-amyloid, and tau loads with hippocampal subregion volume was assessed. The contribution of subregion volumes to the relationship of AD-related deposits on functional cognition was examined with mediation analysis. The effects of AD-related deposits on the rate of subregion atrophy were evaluated with mixed-effects models.

Results

Of 103 participants (mean age: 70.3 years; 37.3% female), 52 had LBD with impaired cognition (LBD-I), 26 had normal cognition (LBD-N), and 25 were A− healthy controls (HCs). Volumes of hippocampal subregions prone to AD copathologies, including subiculum (F = 6.9, p = 0.002), presubiculum (F = 7.3, p = 0.001), and parasubiculum (F = 5.9, p = 0.004), were reduced in LBD-I compared with LBD-N and HC. Volume was preserved in CA2/3, Lewy pathology susceptible subregions. In LBD-I, reduced CA1, subiculum, and presubiculum volumes were associated with greater functional cognitive impairment (all p < 0.05). Compared with HC, subiculum volume was reduced in A+T+ but not A−T− participants (F = 2.62, p = 0.043). Reduced subiculum volume mediated the effect of amyloid on functional cognition (0.12, 95% CI: 0.005 to 0.26, p = 0.040). In 26 longitudinally-evaluated participants, baseline tau deposition was associated with faster CA1 (p = 0.021) and subiculum (p = 0.002) atrophy.

Discussion

In LBD, volume loss in hippocampal output subregions—particularly the subiculum—is associated with functional cognition and AD-related deposits. Tau deposition appears to accelerate subiculum and CA1 atrophy, whereas Aβ does not. Subiculum volume may have value as a biomarker of AD copathology-mediated neurodegeneration and disease progression.

Introduction

The neuropathologic hallmark of the Lewy body diseases (LBD), which include dementia with Lewy bodies (DLB), Parkinson disease (PD), and PD dementia (PDD), is the deposition of aggregated α-synuclein in Lewy bodies and neurites, but comorbid pathologic changes of Alzheimer disease (AD), including amyloid plaques and neurofibrillary tangles, are commonly observed at autopsy.¹ These co-occurring AD pathologies contribute to regional neurodegeneration and volume loss.²

The hippocampus is a consistent site of involvement of both Lewy pathology and AD copathologies.^3,4 Hippocampal neurodegeneration in DLB and PDD is associated with both impairment of short-term memory and hippocampal volume loss compared with healthy control (HC) and participants with nondemented PD,^5,6 albeit with relative preservation compared with AD,⁷ where the burden of AD pathology and neuronal loss is greater.

Much is known about the burden of Lewy and AD-related neuropathologies across hippocampal subregions at autopsy. Although α-synuclein deposition is most prominent in cornu ammonis (CA) 2/3,⁸ amyloid and tau pathologies are most prominent in CA1 and in subiculum and its associated structures.^4,9,10 Thus, volume loss in specific hippocampal subregions may reflect distinct pathologic pathways that contribute to cognitive impairment in LBD. However, most structural imaging studies of the hippocampus have treated it as a single structure, and much less is known about how the differential burden of Lewy and AD copathologies across hippocampal subfields contributes to their volume loss in LBD.

The subfield segmentation tool in FreeSurfer has enabled volume measurements of hippocampal subfields using standard T1-weighted images.¹¹ Using this approach, hippocampal CA1 atrophy has been linked to cognitive impairment in PD.^12,13 Volume loss in CA2/3, CA4/DG, and subiculum have been reported in nondemented PD compared with heathy controls,¹⁴ and early atrophy of CA1, subiculum, and presubiculum has been found to predict conversion to PD dementia.¹⁵ By contrast, in DLB, hippocampal subfield volumes have been reported to be preserved in some studies,^16,17 yet affected in others, including the subiculum, presubiculum, and CA2/3, but sparing CA1.¹⁸

Given the overlap in Lewy pathology and the variability of coincident AD pathologies in PD, PDD, and DLB and the prior demonstration outside of LBD that AD copathologies can lead to hippocampal shape distortions across multiple subregions,¹⁹ here we sought to investigate changes in hippocampal subregion volumes across the LBD spectrum and relate them to cognitive function and PET imaging measures of amyloid and tau. We hypothesized that DLB and PDD would share volume loss in AD-prone subregions and that the degree of volume loss would relate both to the extent of amyloid and tau deposition and to the severity of cognitive impairment. We also hypothesized that in “pure” DLB and PDD without imaging evidence of AD copathologies, α-synuclein pathology might manifest as volume loss in CA2/3.

Methods

Participants

A total of 78 participants with LBD were enrolled from the Massachusetts Alzheimer's Disease Research Center's Memory and Aging Cohort. Of them, 52 had cognitive impairment (LBD-I), comprising 34 who met Lewy Body Consortium criteria for probable DLB²⁰ and 18 cognitively impaired PD (PD-I), including 10 who met Movement Disorder Society (MDS) criteria for probable PDD²¹ and 8 who met level II MDS Task Force guidelines²² for PD with MCI; 26 LBD participants had normal cognitive function (LBD-N) and met PD clinical criteria of the UK Parkinson's Disease Society Brain Bank,²³ and 21 participants (11 with clinically probable DLB) came to autopsy; all had Lewy body disease. All cases in this cohort had motor manifestations of parkinsonism, which is reflected in striatal dopamine deficiency on dopamine imaging. A group of 25 amyloid-negative (A−) HC (mean age: 67.6 ± 4.9, female: 35%) measured with PiB-PET²⁴ were included for comparison.

Neuropsychological Testing

Participants underwent the ADRC's Uniform Data Set neuropsychological tests,²⁵ including the Montreal Cognitive Assessment (MoCA) or the Mini-Mental State Examination (MMSE), as well as the Neuropsychiatric Inventory Questionnaire (NPIQ), and were scored by ADRC physicians on the Clinical Dementia Rating scale Sum-of-Boxes (CDR-SOB). In cases where MoCA scores were unavailable, MMSE scores were converted to MoCA scores.²⁶

Structural Imaging

All participants underwent a structural MRI scan using a Siemens Tim Trio 3T system with a 12-channel phased array head coil as described previously.² T1-weighted images were processed with FreeSurfer to identify the gray-white junction and pial surface and were then manually checked for quality. Participants with evidence of chronic lacunar or territorial infarcts or significant white matter disease (periventricular WMH Fazekas score >2 or deep WMH Fazekas score >1) were excluded. To segment hippocampal subregions, we used FreeSurferV7, which relies on a probabilistic atlas generated with ultra-high-resolution ex vivo MRI data.¹¹

MRI was repeated in 26 participants (13 LBD-I, 13 LBD-N), with interval 0.8–6.6 years (median: 2.4). For this purpose, we implemented the FreeSurfer longitudinal pipeline to process images of each participant.²⁷ We created a template for each participant in the longitudinal processing stream using all MRI time points, following by reconstructions with identical manual quality control. We applied the FreeSurfer longitudinal module to segment hippocampal subregions.²⁸

The automated segmentation tool provided data for the entire hippocampus and several subregions (Figure 1A), and we focused on CA1, CA2/3, CA4/DG, subiculum, presubiculum, and parasubiculum as regions of interest (ROI) (Figure 1B). Visual inspection and ranking of subregion volumes were used to perform quality checks on each scan, where CA1 was expected to be the largest and parasubiculum the smallest. These quality checks did not reveal any abnormalities in the series of scans. Because the left and right hippocampal volumes were strongly correlated (r = 0.88, p < 0.001), bilateral hippocampal volumes were assessed in all analyses. Each hippocampal subregion volume was adjusted for estimated intracranial volume.²⁹ The adjusted hippocampal subregion volumes were used in all analyses.

Figure 1. Hippocampal Subfields Visualized in 3 Dimensions.

(A) The 10 regions available by default using FreeSurfer V7. (B) The 6 regions of interest evaluated in this study. (C) Volumetric differences of hippocampal subfields across groups, displayed as Cohen d effect sizes. HC = healthy control; LBD-I = cognitively impaired LBD; LBD-N = cognitively normal LBD.

PET

Of the 78 participants, 70 completed amyloid imaging^2,30 with PiB-PET: 44 LBD-I (28 DLB, 16 PD-I) and 26 LBD-N participants. Tau PET with flortaucipir (FTP)^2,30 was acquired in 43: 32 LBD-I (20 DLB, 12 PD-I) and 11 LBD-N. The demographics (age, proportion female, and education) of PiB-PET imaged and FTP-PET imaged LBD-I participants were comparable with the overall LBD-I sample (each contrast, p > 0.05). PET data were reconstructed, attenuation-corrected, and assessed for count statistics and motion artifacts, followed by coregistration with MRI and sampling with FreeSurfer-derived ROIs. PiB retention was measured as DVR with cerebellar cortex reference, using a cortical aggregate of frontal, lateral temporal, and retrosplenial cortices (FLR). We stratified participants as amyloid positive or negative (A+/−) using the FLR DVR cutoff value of 1.32.² FTP SUVR was computed with cerebellar white matter reference. Temporal meta-ROI tau burden was quantified using tracer binding in a composite region defined in previous studies.³¹

Tau binding was evaluated in entorhinal cortex, the first location of tau pathology accumulation in AD, and the inferior and middle temporal cortex, neocortical sites associated with early tau spread.⁴ To classify participants as tau positive or negative (T+/−), tau positivity was defined as entorhinal tau levels >1.22, with the additional threshold of inferior temporal tau levels >1.56 or middle temporal tau levels >1.48.³² We compared this classification strategy with autopsy results (requiring NFT Braak stage IV and above for tau positivity) as the gold standard in a subset of 16 participants (eTable 1). The sensitivity and specificity of this strategy was 100% and 83.3%, respectively, with the false-positive cases harboring NFT Braak stage II or III. We applied these criteria in all participants with available FTP PET. A−T− was the largest group (35% in DLB, 66.7% in PD-I, and 91% in PD-N), followed by A+T+ in DLB (30%) and A+T− in PD-I (16.7%) and PD-N (9%). The distribution was comparable with that reported in a recent study³³ (eFigure 1).

Statistical Analysis

Demographic and clinical variables were analyzed using t tests or Kruskal-Wallis H tests for normally and nonparametrically distributed continuous data, respectively, and chi-square or Fisher exact tests for categorical data. Hippocampal subregion volumes were correlated with demographic and clinical data using Pearson (continuous data) and point-biserial (dichotomous data) correlations. Group differences in volumes were analyzed using ANCOVA and Tukey post hoc tests, controlling for age. We followed up primary analyses with subgroup analyses to test whether the results held in each diagnostic group. Based on these results and previous literature, we selected CA1, subiculum, and presubiculum as AD-prone regions and CA2/3 as α-synuclein pathology-enriched for subsequent analysis.

To assess whether hippocampal subregion volumes were associated with cognitive impairment, Spearman partial correlation tests were performed controlling for age. To examine differences in hippocampal subregion volumes among amyloid/tau (AT) groups, ANCOVA was performed, accounting for age as a covariate. To assess whether amyloid or tau levels were associated with hippocampal subregion volumes, Spearman partial correlation tests were performed, controlling for age.

Causal mediation analysis evaluated whether hippocampal subregion volumes mediate the effect of amyloid or tau levels on cognitive impairment measured by MoCA or CDR-SOB. Values were standardized before mediation analysis, with significance estimated using a 95% CI based on 10,000 bootstrap samples. All models were adjusted for age. The mediation analysis was performed using Mediation package.

Longitudinal mixed-effects models were used to analyze how amyloid or tau levels influence subiculum atrophy rate, using backward stepwise elimination for model selection. The modeling incorporated primary fixed-effect predictors of baseline amyloid or tau levels, years in the study, and their interactions, along with fixed covariates of baseline age, sex, and their interactions with years in the study. Random effects were structured to include both intercept and subject-specific temporal trajectories, with residuals examined for normality.

Statistical analyses were performed using R (v4.2.0), with a significance threshold set at p < 0.05. Bonferroni correction was applied to control for multiple comparisons as needed.

Standard Protocol Approvals, Registrations, and Patient Consents

All participants provided informed consent under protocols approved by the Partners Human Research Committee.

Data Availability

Anonymized data will be shared by any qualified investigators in compliance with MADRC data sharing protocol.

Results

Sample Characteristics

Fifty-two participants with LBD-I and 26 participants with LBD-N participated in the study. Age at imaging, sex, and years of education were comparable between groups (Table 1). Cognitive function was worse in LBD-I (CDR-SOB: p < 0.001; MoCA: p < 0.001). Intracranial volume was similar between groups.

Table 1.

Demographics and Clinical Characteristics

	LBD-N	LBD-I
			DLB	PDI
N	26	52	34	18
Age	69.5 (7.1)	72.0 (6.2)	71.7 (5.9)	72.6 (6.9)
Sex, F (%)	9 (34.6)	7 (13.5)	3 (8.8)	4 (22.2)
Education	16.6 (3.0)	16.4 (2.5)	16.2 (2.6)	16.7 (2.4)
CDR-SOB	0.2 (0.3)	5.3 (4.2)*	6.0 (4.4)	3.9 (3.3)
MoCA	26.1 (1.6)	17.6 (6.3)*	16.5 (6.8)	19.8 (4.5)
Parkinsonism, (%)	26 (100)	52 (100)	34 (100)	18 (100)
Visual hallucinations, (%)	1 (3.8)	28 (53.8)	22 (64.7)	6 (33.3)

Abbreviations: CDR-SOB = Clinical Dementia Rating scale Sum-of-Boxes; DLB = dementia with Lewy bodies; LBD-I = Lewy body disease with cognitive impairment; LBD-N = Lewy body disease with normal cognition; MoCA = Montreal Cognitive Assessment; PDI = Parkinson disease with cognitive impairment.

The LBD-I group comprises of the DLB and PDI subgroups. Mean (standard deviations) are shown. *p < 0.05 compared with LBD-N.

The LBD-I subgroups of PD-I and DLB had comparable age at imaging, sex, and years of education (Table 1). CDR-SOB and MoCA scores were comparable between PDI and DLB, with a trend toward greater CDR-SOB score in DLB. Intracranial volume was similar between subgroups.

Volume Loss Was Prominent in Hippocampal Output Subregions of Cognitively Impaired LBD Participants

In the LBD cohort as a whole, increasing age was associated with lower hippocampal volume (r = −0.5, p < 0.001). Age was therefore used as a covariate in subsequent analyses. Hippocampal volume was similar between female and male participants and was not related to years of education. Total hippocampal volume was significantly lower in LBD-I than in LBD-N and HC (F = 6.3, p = 0.003). Compared with LBD-N and HC, the LBD-I group had significantly reduced volumes of subiculum (F = 6.9, p = 0.002; p = 0.012, Bonferroni corrected), presubiculum (F = 7.3, p = 0.001; p = 0.006, Bonferroni corrected), and parasubiculum (F = 5.9, p = 0.004; p = 0.024, Bonferroni corrected) (eTable 2). CA2/3 volume loss was not evident in LBD-I or LBD-N participants (F = 2.9, p = 0.056; p = 0.333, Bonferroni corrected). Cohen d effect sizes are shown in Figure 1C.

Hippocampal total and subregion volumes were comparable in PDI and DLB subgroups (Figure 2; eTable 2). Total hippocampal volume was significantly lower in DLB than PDN and HC (F = 4.2, p = 0.007). Compared with PDN and HC, both PDI and DLB groups had significantly reduced subiculum (F = 4.6, p = 0.004; p = 0.024, Bonferroni corrected) volume (eFigure 2). Presubiculum volume loss was also observed in DLB but not in PDI compared with PDN and HC (F = 4.9, p = 0.003; p = 0.019, Bonferroni corrected).

Figure 2. Hippocampal Subregion Volumes Across the LBD Spectrum.

Volumes are adjusted for intracranial volume. ANCOVA, adjusted for age. Omnibus p values are shown. Post hoc test p-values have been corrected for the 6 subregions assessed. *p < 0.05, **p < 0.01. LBD = Lewy body diseases.

Together, these results demonstrate differential volume loss across hippocampal subregions in LBD.

Volume Loss in Hippocampal Output Structures Was Associated With Functional Cognitive Impairment But Not Hallucinations

Across LBD-I participants, lower volumes of CA1 (rho = 0.3, p = 0.032), subiculum (rho = 0.41, p = 0.003), and presubiculum (rho = 0.45, p < 0.001) correlated with worse performance on the MoCA (Figure 3A). Similarly, lower volumes of CA1 (rho = −0.45, p < 0.001), subiculum (rho = −0.38, p = 0.006), and presubiculum (rho = −0.42, p = 0.002) correlated with worse functional cognition measured with the CDR-SOB (Figure 3B). Similar total hippocampal and subregion volumes were observed between LBD-I participants with and without visual hallucinations (eAppendix 1).

Figure 3. The Association Between Subregion Volume Loss and Cognitive Performance Measured by MoCA (A) and CDR-SOB (B).

Partial Spearman correlations within the combined DLB (purple) and PDI (blue) group are shown. The data of the PDN (green) group are presented for visual inspection only. *p < 0.05, **p < 0.01, ***p < 0.001. CDR-SOB = Clinical Dementia Rating scale Sum-of-Boxes; DLB = dementia with Lewy bodies; MoCA = Montreal Cognitive Assessment; PDI = Parkinson disease with cognitive impairment.

In DLB participants, lower volumes of subiculum (rho = 0.47, p = 0.005) and presubiculum (rho = 0.44, p = 0.009) correlated with lower MoCA scores. Subiculum volume trended toward an association with CDR-SOB (rho = −0.32, p = 0.066) while presubiculum volume was significantly associated (rho = −0.39, p = 0.022).

In PDI participants, no association of subregion volumes and MoCA scores was detected. Even so, lower volumes of CA1 (rho = −0.67, p = 0.003) and subiculum (rho = −0.54, p = 0.027) correlated with higher CDR-SOB scores. Lower presubiculum volume trended toward an association with CDR-SOB (rho = −0.43, p = 0.085).

Amyloid and Tau Status Was Associated With Hippocampal Subregion Volume Loss

AT stratified hippocampal subregion volumes are shown in eTable 3. Omnibus testing of subiculum volume across the AT stratified LBD groups and the A− HC group was significant (ANCOVA, F = 2.62, p = 0.043) (Figure 4A). In post hoc comparisons, A+T+ LBD participants had reduced subiculum volume compared with the A− HC group (p = 0.009) and the A−T− LBD group (p = 0.048). This difference remained significant when analyses were limited to LBD-I participants (F = 2.59, p = 0.047; post hoc contrasts: A+T+ vs A- HC, p = 0.022; A+T+ vs A−T−, p = 0.313). In contrast to A+T+ LBD participants, A−T− LBD (and LBD-I) participants had comparable subiculum volume to A− HC participants. No subiculum volume loss was detected in A+T− or A−T+ groups. Optimal differentiation of A+T+ from A−T− was achieved with a subiculum volume cutoff of 360 mm³, based on an area under the curve of 0.84, yielding 92% specificity and 71% sensitivity (eFigure 3, eAppendix 1).

Figure 4. Effects of Amyloid and Tau (AT) Copathologies on Hippocampal Subregion Volume in LBD.

(A) Hippocampal subregion volume loss stratified by AT status. Omnibus ANCOVA p values are shown. (B) The association of hippocampal subregion volumes with PET-measured cortical amyloid burden. Spearman rho is shown. (C) The association of hippocampal subregion volumes with PET-measured temporal meta-ROI tau burden. *p < 0.05, **p < 0.01, ***p < 0.001.

In contrast to findings in the subiculum, there was no detectable effect of amyloid or tau status on volumes of CA1 or the presubiculum.

Notably, CA2/3 volumes in A−T− LBD-I participants were comparable with A− HC and LBD-N participants (F = 1.18, p = 0.331) (eFigure 4A).

Amyloid Accumulation Was Associated With Hippocampal Subregion Volume Loss

To further explore these findings, we evaluated amyloid and tau accumulation as continuous variables. Across all LBD participants, higher cortical amyloid burden was associated with smaller total hippocampal volume (rho = −0.28, p = 0.018). Higher cortical amyloid was associated with smaller subiculum volume (rho = −0.32, p = 0.008) and a trend toward smaller CA1 volume (rho = −0.23, p = 0.051), with no association with either presubiculum volume (rho = −0.15, p = 0.208) or CA2/3 volume (rho = −0.16, p = 0.183) ((Figure 4B, eFigure 4B).

The results were similar in LBD-I participants. Greater cortical amyloid burden was associated with smaller total hippocampal volume (rho = −0.35, p = 0.021), including smaller subiculum volume (rho = −0.38, p = 0.012) and smaller CA1 volume (rho = −0.32, p = 0.036), but not presubiculum volume (rho = −0.17, p = 0.279).

In the DLB subgroup, greater cortical amyloid deposition was associated with smaller subiculum (rho = −0.39, p = 0.045) volume, but not with CA1 (rho = −0.27, p = 0.177) or presubiculum (rho = −0.02, p = 0.906) volume.

In the PDI subgroup, greater cortical amyloid deposition was associated with smaller CA1 volume (rho = −0.58, p = 0.025) and a trend toward an association with smaller subiculum volume (rho = −0.51, p = 0.051), but not presubiculum volume (rho = −0.42, p = 0.118).

Together, these findings show that amyloid's effects on hippocampal volume loss are subregion specific.

Tau Accumulation Was Associated With Hippocampal Subregion Volume Loss

Across all LBD participants, higher temporal meta-ROI tau burden was associated with smaller total hippocampal volume (rho = −0.5, p < 0.001). Higher temporal meta-ROI tau burden was strongly associated with lower volumes of the subiculum (rho = −0.48, p = 0.001), CA1 (rho = −0.47, p = 0.002), and presubiculum (rho = −0.39, p = 0.011) (Figure 4C), but not CA2/3 (rho = −0.12, p = 0.456) (eFigure 4C).

In LBD-I participants, higher temporal meta-ROI tau burden was associated with subiculum volume loss (rho = −0.37, p = 0.045) and a trend toward lower CA1 volume (rho = −0.34, p = 0.062), but not presubiculum volume (rho = −0.21, p = 0.256).

In the DLB subgroup, temporal meta-ROI tau accumulation was associated with volume loss in subiculum (rho = −0.55, p = 0.015) but not in CA1 (rho = −0.34, p = 0.156) or presubiculum (rho = −0.22, p = 0.356).

In the PDI subgroup, no association of tau uptake with CA1, subiculum, or presubiculum volumes was detected.

Thus, like amyloid, greater cortical tau burden was associated with greater subiculum volume loss in LBD.

Subiculum Volume Loss Mediated the Association of Cortical Amyloid Burden With Functional Cognition

In the LBD-I group, higher cortical amyloid burden was associated with greater functional cognitive impairment reflected in the CDR-SOB (rho = 0.35, p = 0.013) and MoCA score (rho = −0.45, p = 0.002), consistent with previous findings.³⁴

The average causal mediation effects (ACMEs) representing the indirect effect of amyloid burden on CDR-SOB in LBD-I that went through subiculum volume were significant (0.12, 95% CI 0.005–0.26, p = 0.040). The average direct effect (ADE) standing for the direct effect of cortical amyloid on CDR-SOB when controlling for subiculum volume did not reach significance (0.232, 95% CI −0.081 to 0.63, p = 0.168). Thus, the mediation analysis supported the existence of a complete mediation, suggesting that the total effect of greater cortical amyloid burden on higher CDR-SOB scores was explained by reduced subiculum volume (Figure 5). A trending effect of partial mediation was detected for MoCA in the LBD-I group (ACME: −0.133, 95% CI −0.293 to 0.00, p = 0.051; ADE: −0.312, 95% CI −0.595 to −0.07, p = 0.014). No significant mediation effects were detected within the DLB or PDI subgroups.

Figure 5. Mediation Effects of Subiculum in the Relationship Between PET-Measured Neocortical Amyloid Burden and Cognitive Impairment Measured by CDR-SOB.

(Top) Total effect of neocortical amyloid burden and cognitive impairment. (Bottom) Mediation effect by subiculum volume loss. CDR-SOB = Clinical Dementia Rating scale Sum-of-Boxes.

These results suggest that subiculum volume loss is a robust measure of amyloid-mediated neurodegeneration linked to functional cognitive impairment.

Higher Temporal Meta-ROI Tau Uptake at Baseline Accelerated Subiculum Atrophy

We performed an exploratory analysis in a small longitudinal cohort (N = 26) to assess rate of volume loss in the subiculum of LBD-I and LBD-N groups and to examine whether baseline amyloid or tau differentially affected rates of atrophy of the hippocampus or its subregions. The annual rate of volume loss in the total hippocampus and in its subregions was comparable across the LBD-I and LBD-N groups (p > 0.05 for each contrast).

No association was detected between baseline cortical amyloid burden and the rate of atrophy in the hippocampus or its subregions (p > 0.05 for each contrast).

By contrast, the presence of tau (T+) was associated with a faster rate of subiculum atrophy (p = 0.002), with an approximate annual reduction of 7.8 ± 2.2 mm³ for subiculum (Figure 6A). Similarly, greater temporal meta-ROI tau burden at baseline was associated with a steeper rate of atrophy in the subiculum (p = 0.048), with no association in the presubiculum (Figure 6B).

Figure 6. Positive Tau Status and Greater Tau Uptake Levels at Baseline Are Associated With Faster CA1 and Subiculum Atrophy in LBD.

(A) In CA1 and subiculum, but not presubiculum, the tau positive group (red) has a faster rate of atrophy than the tau negative group (gray). Predicted volume and 95% CI are shown. (B) The predictive values of temporal meta-ROI tau SUVR on rate of atrophy reflecting low (1 standard deviation below mean), mean, and high (1 standard deviation above mean) levels are shown. Longitudinal mixed-effects model p values are shown.

In subsequent analyses, we assessed CA1 and presubiculum as well. The presence of tau was associated with a faster rate of CA1 (p = 0.021) but not presubiculum (p > 0.05) atrophy, with an approximate annual reduction of 9.9 ± 3.9 mm³ for CA1 (Figure 6A). Greater temporal meta-ROI tau burden at baseline was associated with a trend toward greater atrophy in CA1 (p = 0.063), but not presubiculum (Figure 6B).

Discussion

Building on the differential vulnerability of distinct hippocampal subregions to Lewy and AD copathologies, we investigated the association between hippocampal subregion volume loss, cognitive measures, and AD-related deposits in LBD. Our findings indicate differential volume loss across hippocampal subregions that is greatest in the subiculum and related structures, associated with functional cognitive impairment, and sensitive to both amyloid and tau pathologies. Subiculum volume differentiated A+T+ from A–T– participants with an AUC of 0.84, suggesting that it may have value in the assessment of AD copathology in LBD. By contrast, despite the consistent involvement of Lewy pathology in CA2/3, we did not detect CA2/3 volume loss in LBD participants who lacked imaging biomarkers of AD copathologies. Volume loss in the subiculum, a hippocampal output structure that is vulnerable to AD pathologies,³⁵ was found to mediate the relationship between cortical amyloid burden and functional cognition in LBD. Furthermore, in longitudinal analyses we found that elevated baseline tau was associated with accelerated volume loss in the subiculum and CA1.

In previous studies, the use of FreeSurferV5.3 for hippocampal segmentation revealed smaller volumes of CA2/3, CA4/DG, subiculum, and presubiculum in individuals with DLB compared with healthy controls.^17,18 Nondemented PD had reduced volume of CA2/3, CA4/DG and subiculum in contrast with heathy controls.¹⁴ However, the accuracy of these findings is uncertain given the known limitations of FreeSurferV5.3.³⁶ Using FreesurferV7, this study provides evidence that LBD participants with impaired cognition harbor reduced subiculum, presubiculum, and parasubiculum volumes. Previous studies have reported similar results in participants with DLB using manual labeling and the hippocampal radial distance technique.^37,38 This finding, which is consistent with recent work using FreesurferV6 that showed that individuals with PD who progressed to dementia had smaller CA1, subiculum, and presubiculum volumes,¹⁵ draws further attention to these major output structures of the hippocampus, situated at a key junction between the hippocampus proper and the entorhinal cortex.

The subiculum in particular is a critical hippocampal subregion that receives input primarily from CA1 and layers II and III of the entorhinal cortex and projects widely to cortical and subcortical regions that regulate memory and emotion-related behaviors. The strong and consistent association between cortical amyloid burden and subiculum atrophy in LBD observed in this study is consistent with previous work demonstrating AD pathologies and neuronal loss in the subiculum in postmortem studies of AD and DLB,^39,40 as well as with previous research reporting a significant link between amyloid deposition measured using principal component analysis and subiculum atrophy,⁴¹ although contradictory results have been reported in studies using CSF to detect amyloid.^12,42 Interestingly, mediation analysis revealed that subiculum volume mediates the association between cortical amyloid burden and functional cognitive impairment, suggesting that global AD-related neurodegeneration is well-reflected in subiculum atrophy in LBD. Thus, subiculum volume may prove useful for tracking AD-associated disease progression in LBD.

In this study, we observed both cross-sectional and longitudinal effects of tau deposition in the hippocampus in LBD. Cross-sectionally, tau burden in the temporal meta-ROI cortices, including entorhinal cortex, was associated with reduced volume of the subiculum, presubiculum, and CA1. Longitudinally, tau deposition in the temporal meta-ROI cortices was associated with a significantly increased rate of atrophy in the subiculum and CA1. Previous research has shown the entorhinal cortex to be one of the earliest brain regions of tau deposition, with tau pathology spreading to other regions including the hippocampus and subiculum.⁴ In DLB brains, tau pathology was most frequently found in the terminal axons of the perforant pathway in the subiculum and CA1, where it correlated with the extent of both neurofibrillary tangles and neuronal loss.^43,44 Thus, the effects of entorhinal tau pathology on the subiculum would be anticipated to reflect propagation and secondary neurodegeneration.

In contrast to volume loss in AD-prone regions, volume loss was not detected in CA2/3, where Lewy neurites are common and often associated with cortical Lewy bodies in DLB⁸ and where Lewy pathologic changes are associated with dementia in PD⁴⁵ and cholinergic degeneration.⁴⁶ Although CA2 is difficult to parcellate with current imaging methods, the preservation of CA2/3 volume reduction in participants with LBD, particularly in those with pure LBD-I lacking amyloid or tau pathologies, is consistent with reports demonstrating resistance of CA2 pyramidal neurons to cell death in preclinical models of hypoxia, epilepsy, and traumatic brain injury, as well as after infusion of α-synuclein fibrils into CA2/3.⁴⁷ In any case, CA2/3 volume as measured here appears to be a poor measure of α-synuclein pathologic burden.

This study has a number of limitations. First, as the design of this study is observational and largely cross-sectional, causal inferences cannot be drawn. Even so, we implemented mediation analysis to test our assumptions of causal inferences, and the results supported our assumptions. In addition, we pursued longitudinal analyses in a subset of the data and found further evidence in support of causal relationships between medial temporal tau and hippocampal subfield neurodegeneration. Second, participant groups were defined on the basis of clinical consensus criteria. However, accuracy using these current criteria is high, including in our experience.⁴⁸ Indeed, all cases who came to autopsy had neuropathologic evidence of LBD. In addition, sample sizes for the DLB and PDI subgroup analyses were modest. Thus, although we observed similar findings in LBD-I and the DLB subgroup, lack of a detected effect of tau uptake on hippocampal subregion volume loss in the PDI subgroup may reflect the limited sample size. In this context, future studies focused on hippocampal subfield volume loss in PDI and LBD more broadly will be informative. Future larger studies will also be needed to evaluate the hypothesis that tau mediates the effects of amyloid on cognition in LBD. Although not the focus of this study, limbic-predominant age-related TDP-43 encephalopathy (LATE) is another common driver of CA1 and subiculum atrophy in aging. Future studies that disentangle the contributions of LATE from those of AD copathologies are needed. Finally, the potential limitations of automated segmentation methods, which may not fully capture the unique anatomic features of hippocampal subregions in each case, may affect the accuracy of our results.⁴⁹ Despite these limitations, we were able to detect amyloid-dependent effects on subiculum that are consistent with the predilection of amyloid pathology to this region.⁵⁰

In summary, the results of this study link volume loss of hippocampal output structures, and in particular the subiculum, to functional cognitive impairment and to amyloid and tau copathologies in LBD. Subiculum volume appears to play a pivotal mediating role in the cross-sectional relationship between cortical amyloid burden and functional cognition. By contrast, medial temporal tau deposition in LBD is identified as a driver of subiculum atrophy over time. Together, these findings provide a framework for future investigations of LBD exploring targeted interventions focused on AD copathologies for the prevention and treatment of cognitive impairment.

Glossary

AD

Alzheimer disease

CA

cornu ammonis

CDR-SOB

Clinical Dementia Rating scale Sum-of-Boxes

DLB

dementia with Lewy bodies

HC

healthy control

LBD

Lewy body disease

LBD-I

Lewy body disease with cognitive impairment

LBD-N

Lewy body disease with normal cognition

MMSE

Mini-Mental State Examination

MoCA

Montreal Cognitive Assessment

PD

Parkinson disease

PDD

PD dementia

Appendix. Authors

Name	Location	Contribution
Rong Ye, PhD	Department of Neurology, Massachusetts General Hospital, Boston; Mass General Institute of Neurodegenerative Disease, Charlestown; Lewy Body Dementia Unit, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Anna E. Goodheart, MD	Department of Neurology, Massachusetts General Hospital, Boston; Mass General Institute of Neurodegenerative Disease, Charlestown; Lewy Body Dementia Unit, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Joseph J. Locascio, PhD	Department of Neurology, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Erin Peterec, BS	Department of Neurology, Massachusetts General Hospital, Boston; Mass General Institute of Neurodegenerative Disease, Charlestown; Lewy Body Dementia Unit, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Michael Properzi, BSc	Department of Neurology, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Emma G. Thibault, BS	Department of Neurology, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Erin Chuba, BS	Department of Neurology, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Keith A Johnson, MD	Department of Neurology, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content
Michael J. Brickhouse, BS	Department of Neurology; Frontotemporal Disorders Unit, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Alexandra Touroutoglou, PhD	Department of Neurology; Frontotemporal Disorders Unit, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content
John H. Growdon, MD	Department of Neurology, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Bradford C. Dickerson, MD	Department of Neurology; Frontotemporal Disorders Unit, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Stephen N. Gomperts, MD, PhD	Department of Neurology, Massachusetts General Hospital, Boston; Mass General Institute of Neurodegenerative Disease, Charlestown; Lewy Body Dementia Unit, Massachusetts General Hospital, Boston	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data

Study Funding

Department of Defense CDMRP/W81XW1810516, NINDS 1R21 NS109833, U01 AG016976-11, NIH P30AG062421, Michael J. Fox Foundation for Parkinson's Research.

Disclosure

R. Ye, A. Goodheart, J. Locascio, E. Peterec, M. Properzi, E. Thibault, E. Chuba, K.A. Johnson, M. Brickhouse, A. Touroutoglou, J.H. Growdon, and B.C. Dickerson report no disclosures relevant to the manuscript; S.N. Gomperts has received funding from NIH (R21 NS109833, U01 AG016976, NIA P30AG062421), the Department of Defense CDMRP/W81XW1810516, and the Michael J. Fox Foundation for Parkinson's Research. Go to Neurology.org/N for full disclosures.