Correlating hippocampal and amygdala volumes with neuropathological burden in Down syndrome and Alzheimer’s disease and related neurodegenerative pathologies using 7T postmortem MRI

Abstract

Limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC), is common in elderly brains and often seen in conjunction with Alzheimer’s disease neuropathologic change (ADNC). LATE-NC typically begins in the amygdala and spreads to the hippocampus and neocortex. Whether it contributes to hippocampal and amygdala atrophy in Down syndrome (DS) remains unexplored. We analyzed amygdala and hippocampal volumes and neuropathological burden in 12 DS cases and 54 non-DS cases with AD and related neurodegenerative pathologies (ADRNP) using 7 Tesla (7T) postmortem ex vivo MRI. Postmortem and antemortem hippocampal volumes were significantly correlated in a subset of 17 cases with available antemortem MRI scans. DS cases had smaller hippocampal and amygdala volumes than ADRNP cases; these correlated with more severe Braak stage but not with Thal phase. LATE-NC and hippocampal sclerosis (HS) were uncommon in DS cases. In ADRNP cases, lower hippocampal volumes associated with dementia duration, advanced Thal phase, Braak NFT stage, C score, LATE-NC stage, HS and arteriolosclerosis severity; reduced amygdala volumes correlated with severe LATE-NC stage, HS, and arteriolosclerosis severity, but not with Thal phase or Braak NFT stage. Lewy body pathology did not affect hippocampal or amygdala volume in either cohort. Thus, hippocampal volumes in ADRNP were influenced by both ADNC and LATE-NC, and amygdala volumes were primarily influenced by LATE-NC. In DS, hippocampal and amygdala volumes were primarily influenced by tau pathology.

INTRODUCTION

Aging adults with Down syndrome (DS) accumulate Alzheimer’s disease (AD) neuropathology, including β-amyloid plaques and neurofibrillary tangles, by the age of 40 years.^1,2 By age 56, over half of older individuals with DS develop mild cognitive impairment or early signs of dementia^3–10 with nearly all ultimately dying with AD.¹¹ This prevalence of AD in people with DS limits further increases in life expectancy, underscoring the critical importance of the research for this population and, by extension, the broader AD population. To investigate the underlying pathological mechanisms, a comparative analysis of DS and late-onset AD cohorts will be beneficial.

Limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC) is a common concurrent neurodegenerative pathology in AD brains.¹² It typically begins in the amygdala before spreading to the hippocampus and neocortex. In radiology-to-pathology correlation studies, LATE-NC at the time of autopsy was associated with smaller amygdala¹³ and hippocampal¹⁴ volumes at the time of prior antemortem MRI. Whether similar associations exist in DS, the most common genetic cause of AD, remains unexplored. Investigating the underlying pathologies contributing to hippocampal and amygdala atrophy in DS, especially in a considerably younger population, could offer insights into early AD-related changes in this demographic.

Postmortem imaging offers a unique opportunity to examine 3D radiological images and investigate their underlying pathobiology.^15–22 High-resolution and high-contrast 7 tesla (7T) postmortem MRI can capture more subtle pathological changes including lesions, microinfarcts, and enlarged perivascular spaces as compared to the lower field strengths.^23–25 While postmortem in situ imaging²⁶ allows automated image processing, logistical challenges often make it infeasible. Postmortem ex vivo imaging has been approached in various ways, with some studies dissecting the brain into parts²⁷ or attempting extremely long acquisition times.²⁸ However, these approaches are impractical for larger cohorts. Researchers who conducted postmortem imaging directly in fluids^{23–25,  27–30} or without any embedding medium³¹ may encounter challenges related to immobilizing the tissue during imaging and subsequently aligning the images. Research has shown that agar gel, in contrast to formalin and phosphate-buffered saline, offers superior signal-to-noise and contrast-to-noise ratios as an embedding medium, while also providing stability to the brain.^32,33 We have designed a 3D-printed brain container with coronal and axial cutting guides^34–36 and developed a standardized protocol for brain embedding, postmortem imaging, image processing, and co-registration to block face photographs and histopathological images.

In this study, we present our 7T postmortem MRI methods for whole hemisphere imaging in agar gel and explore the potential correlation of amygdala and hippocampal atrophy with neuropathological burden in DS and a cohort of brains with AD and related neurodegenerative pathologies (ADRNP). A comparative analysis of DS and ADRNP cohorts will help determine whether these observations are associated with Alzheimer’s disease neuropathologic change (ADNC), LATE-NC, Lewy body (LB) pathology, vascular pathology or a combination thereof.

METHODS

Participants

All brain donors were participants in the University of Pittsburgh’s (Pitt) Alzheimer’s Disease Research Center (ADRC), the Ginkgo Evaluation of Memory Study³⁷ long-term follow-up cohort, local brain donors to the Pitt Neurodegenerative Brain Bank or part of the Alzheimer Biomarkers Consortium-Down syndrome (ABC-DS).³⁸

The research protocol has been approved by the Committee for Oversight of Research and Clinical Training Involving Decedents at the University of Pittsburgh and the Institutional Review Board at the University of California Irvine (UCI). This study encompasses postmortem brains obtained between December 2019 and May 2023. Table 1 outlines the fundamental demographics, including age, sex, and race as well as ApoE4 carriership and last clinical diagnosis. The cohort comprises 12 individuals with DS and 54 with ADRNP. Within the ADRNP group, there were 16 cases of AD only, 30 cases of AD + Dementia with Lewy bodies (DLB), 2 cases of AD + argyrophilic grain disease (AGD), 2 cases of AD + DLB + AGD, 1 case of AD + Parkinson disease with dementia, 1 case of DLB only, and 2 cases of primary age-related tauopathy (PART) only. For DS cases, the most recent DS mental status examination (DSMSE) score (range 0-81) is available for 10 of the 12 individuals, with the examination conducted 2.0 ± 1.1 years before autopsy. In the ADRNP group, 44 of the 54 individuals had their most recent mini mental state examination (MMSE) scores (range 0-30) recorded 3.0 ± 2.6 years before death. No cognitive testing information was available for the remainder of the cohort.

Postmortem brain preparation

For postmortem imaging, a reusable brain container that fully fits 1 supratentorial hemisphere without cerebellum or brainstem has been designed and printed.^{36,  39} The container consists of 4 components: a screw cap, domed lid, a cutting guide for coronal slabs, and a container base. Utilizing a Fortus 450 3D printer (Stratasys, Eden Prairie, MN, USA), these containers were printed using high-density filament polycarbonate (Stratasys, USA). The cutting guides, spaced 0.5 mm apart, facilitate precise brain cutting of coronal slabs at 1 cm thickness by neuropathologists.

The average postmortem interval for this cohort was 10.5 hours in cases with DS and 8.7 hours in those with ADRNP (Table 2). Brains were weighed fresh at the time of autopsy prior to dissection. The brainstem was separated at the midbrain level, and the forebrain hemispheres were bisected. The left hemisphere was then fixed in 10% formalin (ADRNP cases + subset of DS cases), or 4% paraformaldehyde (remaining DS cases) for a minimum of 3 weeks. To prepare the left hemispheres for imaging, the leptomeninges were removed from the cortical surface to minimize air bubble entrapment. We employed a mixture of 1.5% (w/v) agar (A5431, Millipore Sigma, Burlington, MA, USA) and 30% sucrose (S5-3, Fisher Chemical, Pittsburgh, PA, USA) to embed the brain in our 3D-printed container. While submerging the brain in the container, we gently massaged it to release any remaining trapped air bubbles. Once the agar was completely solidified, the container was sealed with a lid and placed inside the scanner's head coil for imaging.

Neuroimaging

Postmortem scans were performed using a 7T human MRI scanner (Siemens Magnetom, Forchheim, Germany) with both first^40–42 and second^43,44 generation of Tic-Tac-Toe head coil radiofrequency system. Structural imaging included acquisition of T1-weighted (T1w) MP2RAGE images at a resolution of 0.37 mm, T2-weighted (T2w) SPACE images at a resolution of 0.41 mm, and T2-star (T2*) GRE images at a resolution of 0.37 mm, with detailed sequence parameters provided in Table 3. Ex vivo T1w and T2w images were registered by rigid registration, and manual segmentation of the hippocampus and the amygdala was performed using ITK-SNAP.⁴⁵ To ensure accuracy, all segmentations were evaluated by 2 authors (J.J.L. and J.N.) before exporting the raw volumes.

Antemortem MR scans were available for 1 DS and 16 ADRNP cases and were retrieved from various studies and clinical radiology records. These scans utilized either a 3T Siemens scanner (Siemens, Munich, Germany) or a 3 T GE scanner (GE Healthcare, Chicago, IL, USA). T1w MPRAGE images were acquired with isotropic resolutions ranging from 0.8 to 1.2 mm and acquisition times ranging between 4 and 9 minutes. For antemortem T1w images, hippocampus and amygdala segmentation was performed using FreeSurfer (version 7.3.1, http://surfer.nmr.mgh.harvard.edu/). The extracted antemortem hippocampal and amygdala volumes were correlated with the ex vivo volumes, accounting for discrepancies in absolute values due to the scan interval (between antemortem MRI and postmortem ex vivo MRI) and tissue deformation during fixation.

Neuropathology

Tissue sampling and staining included all brain regions recommended by the 2012 National Institutes of Aging—Alzheimer’s Association (NIA-AA) consensus criteria for the neuropathological evaluation of Alzheimer’s disease.^46,47 Immunohistochemical staining for β-amyloid was performed to generate Thal phases.⁴⁸ Phospho-Tau (p-Tau) staining was performed to determine Braak NFT stage.⁴⁹ Modified Bielschowsky stains were used to assess neuritic plaque density by Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) criteria⁵⁰ for the ADRNP cohort and DS cases evaluated at Pitt. In the DS cases from UCI, neuritic plaque scores were based on p-Tau stains. All cases were assigned ABC scores following the NIA-AA criteria, with details of antibodies provided in Table 4.

TDP‐43 immunohistochemistry was carried out on sections from the amygdala, hippocampus‐mesial temporal cortex, and midfrontal neocortical regions following consensus guidelines.¹² TDP-43-positive cases were given a diagnosis of LATE-NC. Severity of LATE-NC pathology was assessed by 2 different methodologies. First, LATE-NC stage was determined following published guidelines¹² by classifying cases based on brain region involvement into stage 1 (amygdala only), stage 2 (stage 1 + hippocampus and/or entorhinal/transentorhinal cortex), and stage 3 (stage 2 + midfrontal cortex). In addition, the severity of TDP-43 pathology in the following 5 regions was assessed on a semiquantitative scale (none = 0, mild = 1, moderate = 2, severe = 3): amygdala, CA1, dentate gyrus, entorhinal/transentorhinal cortex and midfrontal cortex. Stage 2 regions (CA1, dentate gyrus and mesial temporal cortex) were averaged before combining with amygdala and midfrontal cortex score for a final severity score (ranging from 0 to 9). Given the low frequency of TDP-43 pathology in the DS cases, severity scores were only obtained for the ADRNP cohort.

Hippocampal sclerosis (HS) was evaluated unilaterally in the left hemisphere at two coronal sections, one from the anterior hippocampus and the other from the mid‐hippocampus at the level of the lateral geniculate body. The presence of HS was determined based on severe neuronal loss and gliosis in CA1 and/or subiculum, disproportionate to ADNC in the same regions, while blinded to TDP-43 pathology status.^{46,  51}

LB stage included stage 0 (negative), stage 1 (olfactory bulb only or brainstem-predominant), stage 2 (amygdala-predominant), stage 3 (limbic), and stage 4 (diffuse neocortical), according to published studies^52–54 and the fourth consensus report of the DLB consortium.⁵⁵

Assessment of cerebrovascular pathologies included arteriolosclerosis evaluated in the white matter, atherosclerosis assessed in basal cerebral arteries, and cerebral amyloid angiopathy (CAA). All cerebrovascular pathologies were scored on a semiquantitative scale as 0 = absent, 1 = mild, 2 = moderate, 3 = severe.

Statistical analysis

All statistical analyses were conducted using Prism (version 10.2.0.392) unless specified otherwise. Hippocampal and amygdala volumes were normalized to fresh brain weight and adjusted for age, sex, and ApoE4 carriership using MATLAB (version R2023a). D'Agostino-Pearson normality tests were performed to assess whether values were normally distributed. If normal, subsequent Pearson correlation was employed to assess relationships between volume and neuropathological burden with significance set at P < .05; if not normal, Spearman nonparametric correlation was performed to assess significance. Pearson or Spearman correlation coefficients were reported when P < .05. Unpaired t-tests were utilized to compare 2 different groups.

Stepwise regression was conducted using SPSS (version 29.0.2.0) to explore which independent variables contribute the most to postmortem hippocampal and amygdala volumes. The dependent variable was the hippocampal or amygdala volume normalized to fresh brain weight and adjusted for age, sex, and ApoE4; the eight independent variables included Thal phase, Braak NFT stage, C score, LATE-NC stage, LB stage, arteriolosclerosis, atherosclerosis, and CAA. The resulting R², unstandardized coefficients, and associated P-values are reported.

RESULTS

Postmortem-antemortem volume correlation

The T1w and T2w images of each brain were registered and used for hippocampal and amygdala segmentation. Segmentations were created manually and quality-inspected (Figure 1A). Among all brains, 17 had antemortem scans with an average antemortem-postmortem scan interval of 4.6 ± 4.1 years [range 0-13 years]. Pearson correlation analysis revealed a significant positive correlation between the postmortem hippocampal volume and that of the antemortem scans (Pearson r = 0.6307, P = .0066) (Figure 1B). While a similar trend was observed for the amygdala, it did not reach significance (P = .1182) (Figure 1C). The volume difference between postmortem and antemortem scans was inversely correlated with the scan interval in the hippocampus (Spearman r = −0.6026, P = .0095) (Figure 1D) but not in the amygdala (P = .0968) (Figure 1E). It is worth noting that there was one LATE-NC case with a long scan interval of 11 years and a dramatic decrease of amygdala volume between antemortem and postmortem scans (from 2072 to 638 mm³). This outlier likely explains the worse correlation for amygdala compared to hippocampal volumes. Overall, these findings indicate that despite the absence of intracranial volume or whole hemisphere volume, the postmortem hippocampal volume derived from manual segmentation positively correlates with the antemortem hippocampal volume.

DS-ADRNP volume comparison

The average age of individuals with DS was ∼20 years younger than those with ADRNP, with more males and exclusively white participants (Table 1). Regarding ApoE4 carrier status, the majority of DS participants were non-carriers (67%), whereas almost an equal number of carriers (39%) and non-carriers (30%) were reported in the ADRNP cohort. We were unable to obtain postmortem amygdala volumes in 4 ADRNP cases due to poor delineation in the earliest MRI scans when the imaging protocol was still under development, resulting in an ADRNP sample size of 50 for all amygdala volume analyses. Increased attention to removing entrapped intraventricular air bubbles in the temporal horn improved amygdala delineation in subsequent MRI scans. When comparing hippocampal and amygdala volumes, individuals with DS had significantly lower hippocampal (P = .0004, Figure 2A) and amygdala volumes (P = .0002; Figure 2B) compared to ADRNP cases, after adjusting for age, sex, and ApoE4 carriership.

DS and ADRNP volume-clinical correlation

In DS, the most recent DSMSE was obtained on average 2 years prior to autopsy and most DS participants had already developed dementia (75%) based on the last clinical diagnosis; however, we had insufficient data to calculate dementia duration for this cohort. In DS, no correlation between DSMSE and hippocampal volume was detected (P = .2233) (Figure 3A), but a trend towards significance was observed in the amygdala (P = .0723) (Figure 3B).

In the ADRNP cohort, 67% were clinically diagnosed with probable AD, 18% with other forms of dementia, 9% with mild cognitive impairment, and 6% had no cognitive impairment at the last clinical visit. The duration of cognitive impairment was 11.1 ± 4.3 years. In ADRNP, the hippocampal volume had an inverse correlation with dementia duration (Pearson r = −0.3996, P = .0174), after adjusting for age, sex, and ApoE4 carriership (Figure 3C). However, this correlation did not reach significance in the amygdala (P = .2086) (Figure 3D). The last MMSE, obtained 3.0 ± 2.6 years prior to autopsy, did not show a significant correlation with hippocampal (P = .5826; Figure 3E) or amygdala (P = .7548; Figure 3F) volume. No correlation between the last MMSE score and the MMSE-autopsy interval was detected (P = .2597).

Neuropathological burden in DS

The distribution of Thal phase, Braak NFT stage, C score, LATE-NC stage, presence of HS, LB stage, atherosclerosis, arteriolosclerosis, and CAA, is summarized in Table 2. In the DS group, no significant correlations were detected between hippocampal volume and Thal phase (P = .9325) or C score (P = .1032) (Figure 4A and C), but lower hippocampal volume correlated with advanced Braak NFT stage (Pearson r = −0.7481, P = .0051) (Figure 4B). In the 2 DS cases that exhibited LATE-NC pathology (P = .2838) (Figure 4D) or HS (P = .2221) (Figure 4E), trends toward smaller hippocampal volumes were observed but did not reach significance. In the 4 DS cases that exhibited LB pathology, 2 with amygdala-predominant LB and 2 with diffuse neocortical LB, no significant difference in hippocampal volume (P = .5262) (Figure 4F) was observed when compared to the non-LB group. Regarding cerebrovascular pathologies, no atherosclerosis was present in our DS cohort. No significant correlation of hippocampal volume was detected with the severity of CAA (P = .4694) or arteriolosclerosis (P = .4300) (Figure 4G and H).

Similarly, no significant correlation of amygdala volume was detected with Thal phase (P = .9921), C score (P = .0682), CAA (P = .2263), or arteriolosclerosis (P = .4357) (Figure 5A, C, G, and H). The amygdala volume was correlated with advanced Braak NFT stage (Pearson r = −0.6915, P = .0127) (Figure 5B). No significant difference was detected in amygdala volume between cases with and without LATE-NC (P = .1610), HS (P = .3913), or LB (P = .9124) (Figure 5D-F).

Neuropathological burden in ADRNP

In ADRNP cases, lower hippocampal volume correlated with more advanced Thal phase (Spearman r = −0.3406, P = .0117), Braak NFT stage (Spearman r = −0.2838, P = .0376), C score (Spearman r = −0.4422, P = .0008), LATE-NC stage (Spearman r = −0.4271, P = .0013), arteriolosclerosis severity (Pearson r = −0.3003, P = .0274), and was associated with the presence of HS (P < .0001) (Figure 6A-E and H). No significant correlation of hippocampal volume was detected with LB stage (P = .9707), CAA (P = .1678), or atherosclerosis (P = .2749) (Figure 6F, G, and I).

No significant correlation of amygdala volume with Thal phase (P = .2454), Braak NFT stage (P = .9749), C score (P = .8934), LB stage (P = .3945), CAA (P = .1984), or atherosclerosis (P = .0899) was detected (Figure 7A-C, F, G, and I). Lower amygdala volume correlated with more advanced LATE-NC stage (Spearman r = −0.3348, P = .0175), presence of HS (P = .0139), and arteriolosclerosis severity (Pearson r = −0.3270, P = .0205) (Figure 7D, E, and H).

We conducted stepwise regression analyses to determine the contribution of the 8 independent variables (Thal phase, Braak NFT stage, C score, LATE-NC stage, LB stage, atherosclerosis, arteriolosclerosis, and CAA) to the variance in hippocampal and amygdala volumes, which were normalized to fresh brain weight and adjusted for age, sex, and ApoE4 status in the ADRNP cohort. This stepwise regression revealed that C score and LATE-NC stage were the predictors in postmortem hippocampal volume (Figure 8A). The C score accounted for 18.3% of the variance, followed by the LATE-NC stage with an additional 8.8%, totaling 27.1% of the variance in postmortem hippocampal volume. In the case of postmortem amygdala volume (Figure 8B), LATE-NC (9.6%) and arteriolosclerosis severity (6.3%) emerged as the predictors in the stepwise regression model, accounting for a total of 15.9% of the variance.

DISCUSSION

We implemented a postmortem ex vivo MRI protocol to investigate the correlation between hippocampal and amygdala volume and neuropathological burden in neurodegenerative diseases. The significant correlation found between antemortem and postmortem hippocampal volumes suggests the reliability of manually segmented volumes, particularly in ex-vivo scanned cases where intracranial or whole hemisphere volume is unavailable. While we observed a trend towards correlation between antemortem and postmortem volumes in the amygdala, this correlation did not reach significance, likely due to two primary reasons. First, delineation of anatomic structures in the anterior medial temporal lobe was challenging in four early MRI scans, while we were still refining the imaging protocol. Second, an outlier case with a scan interval of 11 years and severe amygdala volume loss over the years may have contributed to this lack of statistical significance.

In our analysis, we observed significantly smaller hippocampal and amygdala volumes in individuals with DS compared to those in the ADRNP cohort, which is consistent with findings from previous studies.^56–61 Interestingly, research on children with DS revealed no differences in amygdala volumes between DS and control groups but hippocampal volumes were notably smaller in the DS group.⁶² This suggests that decreased hippocampal volumes in DS may also be attributed to early developmental differences in addition to neurodegenerative changes. Hippocampal volumes have been shown to correlate with cognitive measures (DSMSE scores) in DS cases.⁵⁹ While we observed trends in the same direction for both hippocampal and amygdala volumes, our analyses did not reach significance, likely due to the overall low number of cases in our DS cohort. In contrast, within our ADNRP cohort, hippocampal volume exhibited an inverse correlation with dementia duration, a novel finding in our study and consistent with our previously reported association of HS and LATE-NC pathology with duration of cognitive symptoms in a non-overlapping cohort of cases.⁶³ However, we did not find a significant correlation between hippocampal or amygdala volume and the last MMSE score, even after adjusting for the MMSE-autopsy interval, contrary to findings in some prior studies,^64–68 and likely due to differences in sample size and group ratios.

Our neuropathological results showed that both hippocampal and amygdala volumes correlated with Braak NFT stage in the DS cohort, after adjusting for age, sex, and ApoE4 carriership. However, no correlations with Thal phase were detected. While LATE-NC pathology was not common in the DS cohort, its frequency was consistent with findings from other studies in younger populations that reported LATE-NC pathology in 9%-18% of DS, familial and early-onset sporadic AD cases.^69–73 Despite the low numbers, we observed a trend towards small hippocampal and amygdala volumes in the DS cases with LATE-NC. Similar trends were observed for the two DS cases that exhibited HS. Interestingly, both HS cases in this cohort were negative for TDP-43 pathology, which is consistent with other studies where half of the DS cases with HS were non-LATE-NC cases.^70,71 One of our cases had a history of epilepsy with onset concurrent with dementia. Neither case had a known history of cardiac arrest or cardiac malformation to explain the severe neuronal loss in the hippocampus. Thus, the HS in these cases is best explained by the presence of ADNC. We also identified 4 DS cases with LB pathology (34%), which falls within the published prevalence range of 8%-50%.^74,75

In our ADRNP cohort, our neuropathological findings are consistent with previous research, demonstrating an inverse correlation between hippocampal volume and the Thal phase,^76,77 Braak NFT stage,^78–80  C score^81–83, and LATE-NC stage.^{65,  77,  84,85} We found that the C score accounted for 18.3% of the variance in postmortem hippocampal volume. This finding is consistent with a previous study which showed that the significant correlation between hippocampal volume and the number of neurons in the hippocampal subfield CA1 may explain the hippocampal atrophy observed in severe cases with neuritic plaques.⁸¹ Additionally, a study suggested that including the TDP-43 stage could explain ∼3% of the variance in hippocampal volume.⁸⁵ In our stepwise regression analysis, the LATE-NC stage accounts for 8.8% of the variance in the hippocampal volume and 9.6% of the variance in amygdala volume. Consistent with previous findings,^86–88 we observed lower hippocampal volumes in cases with compared to those without HS. Interestingly, TDP-43/HS pathology seems to have a stronger association with hippocampal volume than ADNC in older individuals with AD.⁸⁵ Compared to that study with a mean age at death of around 90 years, the average age in our ADRNP cohort was about a decade younger, which may explain why ADNC and LATE-NC pathologies had comparable effect sizes in our cohort. Furthermore, we noted an inverse correlation between amygdala volume and LATE-NC stage. In our stepwise regression analysis, the LATE-NC stage and arteriolosclerosis emerged as the predictors for postmortem amygdala volume with the latter contributing to 6.3% of the variance. Studies have indicated that the amygdala is vulnerable to arteriolosclerosis^89–91 but the frequency of arteriolosclerosis does not differ by TDP-43 stage.⁹² Regarding the contribution of LB, antemortem imaging studies have reported that less severe hippocampal atrophy is associated with DLB when compared to AD.^93,94 We did not observe any significant differences in hippocampal or amygdala volume between LB and non-LB cases, nor did we find an association between volumes and LB stage. These results align with previous findings of comparable amygdala volumes between AD and DLB.⁶⁸

This exploratory study is limited by the small sample size, particularly affecting analyses of the DS cohort. Previous studies have indicated that female individuals with DS tend to experience longer durations of dementia⁹⁵ and exhibit higher tau burden in vivo,⁹⁶ while males with DS and the ApoE4 allele display elevated white matter hyperintensity volumes in the occipital lobe.⁹⁷ Due to the small sample size of our DS cohort, we had to defer exploration of sex differences in neuropathological burden to follow-up studies when additional cases in the DS cohort will have come to autopsy. Despite this limitation, our study provides novel insights into differential impact of neurodegenerative pathologies on regional volumes between DS and sporadic AD cases. We acknowledge that neuritic plaques are not completely visualized with p-Tau immunohistochemistry and that the neuropathology assessment in neuritic plaques is different between Pitt (ADRNP + subset of DS cases) and UCI (remaining DS cases). Antemortem MRI scans were only available in a subset of cases and were obtained using different scanners and scanning protocols preventing us from performing comparative analyses with postmortem neuropathology burden. Given the small samples size, we did not account for the variability in scanners and scanning protocols or perform multi-scanner harmonization. Lastly, we only analyzed overall severity scores for ADNC but did not assess regional disease burden for Aβ or p-Tau pathologies in the hippocampus or amygdala.

CONCLUSIONS

Our 7 T postmortem MRI protocol produced good alignment of postmortem volumes with antemortem findings and revealed correlations between volume measures and neuropathological burden. In ADRNP, hippocampal volume is influenced by both ADNC and LATE-NC, whereas amygdala volume appears to be influenced primarily by LATE-NC. In DS, hippocampal and amygdala volumes are smaller than in ADRNP and primarily influenced by tau pathology. Our high-resolution, high-contrast 7T MRI protocol presents an opportunity that lays the foundation for future studies into white matter pathologies and targeted tissue sampling.

Acknowledgments

Data used in preparation of this article were obtained from the Neurodegeneration in Aging Down syndrome (NiAD) database (niad.loni.usc.edu) and Alzheimer’s Disease in Down syndrome (ADDS) database. As such, the investigators within the ABC-DS study contributed to the design and implementation of ABC-DS and/or provided data but did not participate in analysis or writing of this report. A complete listing of ABC-DS investigators can be found at: https://www.nia.nih.gov/research/abc-ds#data The authors express their sincere gratitude to all patients and their dedicated service providers and families who generously volunteered as participants in these studies. Additionally, the authors extend their heartfelt thanks to the staff members who helped with autopsy coordination and study administration, including Sierra Wright, Joni Vander Bilt, and the autopsy coordination and brain banking teams of the Pitt ADRC. Special appreciation is also extended to the other principal investigators at the University of Pittsburgh for their collaboration and for sharing of antemortem scans with the research team.

Alzheimer Biomarker Consortium-Down Syndrome (ABC-DS) Investigators: Beau M. Ances, MD, Howard F. Andrews, PhD; Karen Bell, PhD; Rasmus M. Birn, MD; Adam M. Brickman, PhD; Peter Bulova, PhD; Amrita Cheema, MD; Kewei Chen, PhD; Bradley T. Christian, PhD; Isabel Clare, PhD; Lorraine Clark, PhD; Ann D. Cohen, PhD; John N. Constantino, PhD; Eric W. Doran, MD; Anne Fagan, MS; Eleanor Feingold, PhD; Tatiana M. Foroud, PhD; Benjamin L. Handen, PhD; Sigan L. Hartley, PhD; Rachel Henson, PhD; Christy Hom, PhD; Lawrence Honig, PhD; Sterling C Johnson, MD; Courtney Jordan, PhD; RN; M. Ilyas Kamboh, PhD; David Keator, PhD; William E. Klunk, MD PhD; William Charles Kreisl, MD; Sharon J. Krinsky-McHale, PhD; Florence Lai, MD; Patrick Lao, PhD; Charles Laymon, PhD; Joseph Hyungwoo Lee, PhD; Ira T. Lott, MD; Victoria Lupson, PhD; Mark Mapstone, PhD; Chester A. Mathis, PhD; Davneet Singh Minhas, PhD; Neelesh Nadkarni, MD; Sid O'Bryant, PhD; Deborah Pang, MPH; Melissa Petersen, PhD; Julie C. Price, PhD; Margaret Pulsifer, PhD; Michael Rafii, MD, Eric Reiman, PhD; Batool Rizvi, MD; Herminia Diana Rosas, MS; Marwan N. Sabbagh, MD; Nicole Schupf, MD; Wayne P. Silverman, PhD; Dana L. Tudorascu, PhD; Rameshwari Tumuluru, PhD; Benjamin Tycko, MD; MD, Badri Varadarajan, PhD; Desiree A. White, PhD; Michael A. Yassa, PhD; Shahid Zaman, PhD; MD, Fan Zhang PhD; PhD.

Funding

Data collection and sharing for this project was supported by the National Institutes of Health R01AG063525 (T.S.I.), R01MH111265 (T.S.I.), P30AG066468 (H.J.A., M.D.I., and J.K.K.), R01AG069912 (J.K.K.), U19AG068054 (E.H., M.D.I., T.S.I., and J.K.K.), P30AG066519 (E.H.) and the Alzheimer Disease Biomarker Consortium on Down syndrome and the Eunice Kennedy Shriver National Institute of Child Health and Human Development U01AG051406 (E.H.) and U01AG051412 (E.H.), as well as the Bioengineering in Psychiatry Training Program T32MH119168 (J.B.). This research was supported in part by the University of Pittsburgh Center for Research Computing, RRID: SCR_022735, through the resources provided. Specifically, this work used the H2P cluster, which is supported by NSF award number OAC-2117681.

Conflicts of interest

The authors declare that they have no competing interests.

Data availability

All data generated and analyzed during this study are included in this article. The anonymized raw and processed MRI scans, along with subject-level and group-level hippocampus and amygdala segmentations and neuropathology data, are available upon reasonable request.

Ethics approval and consent to participate

The research protocol has been approved by the Committee for Oversight of Research and Clinical Training Involving Decedents at the University of Pittsburgh and by the Institutional Review Board at the University of California, Irvine.

References

Author notes