Background

Alzheimer’s disease (AD) is a devastating disease associated with brain deposition of amyloid-β (Aβ) and tau, and irreversible neurodegeneration. Aβ deposition occurs early in the AD process, decades before cognitive decline1. In late-onset AD, Aβ accumulation is due to failed clearance – not overproduction2,3. The strongest genetic risk factor for AD, the ε4 allele of the APOE gene4,5,6, is known to impair the rate of Aβ clearance7. However, the mechanisms of clearance failure in AD and in association with the APOE4 allele remain uncertain. As detailed in several comprehensive reviews3,8,9,10, Aβ is normally removed from the brain by three main mechanisms: (1) fluid clearance including the recently described glymphatic system, (2) transport or degradation at the endothelial blood-brain barrier, and (3) transport across the blood-CSF barrier, which is the choroid plexus (CP). Of these mechanisms, the role of CP in Aβ clearance has received less research attention and is the focus of this study.

Until recently, the main function of the CP was considered to be production of CSF through blood filtration. Now, the CP is understood to play a critical role in maintaining brain homeostasis by clearing toxins and metabolic waste including Aβ11,12,13,14,15,16,17,18,19. CP also regulates inflammation20,21,22 and have been shown to play a role in neurogenesis23. All of these processes are critical to AD pathogenesis.

Because CP dysfunction may be an early event in the pathologic cascade leading to AD15,16, in vivo identification of CP abnormalities using neuroimaging could potentially identify individuals at risk of AD before any involvement of brain parenchyma. This has motivated several recent studies measuring CP volume using MRI. Intriguingly, these studies have consistently demonstrated larger CP in patients with AD compared to cognitively normal controls24,25,26 and in association with the severity of cognitive impairment24,25,27. CP enlargement is also associated with older age in subjects with and without AD24,25,27,28. In one of the only longitudinal studies of CP structural change, CP enlargement with aging was found to be significantly greater in individuals who experienced cognitive decline compared to those who remained cognitively stable24. Importantly, CP enlargement was also strongly associated with APOE4 status, with APOE4 homozygotes showing a threefold faster rate of CP enlargement compared to heterozygotes and non-carriers24.

Pathologic CP enlargement contrasts with atrophy of virtually all brain parenchymal structures with aging and AD. The basis of pathologic CP enlargement in human neuroimaging studies of AD is uncertain. We recently showed that the volume of calcium within CP - measured using a multimodal Computed Tomography (CT) and MRI method - contributes to CP overall volume measured using MRI20. In that study, the volume of CP calcium but not CP overall volume, was associated with PET-measured cortical inflammation in healthy subjects. CP calcification is well known to increase with normal aging29,30 and has been associated with neurodegeneration31,32. Our prior finding20 suggested that the volume of calcium deposited within CP might be a more accurate/sensitive biomarker of CP dysfunction than CP overall volume, at least with respect to the role of CP in regulation of inflammation. It is unknown whether this may also be the case with respect to the role of CP in clearing Aβ and preventing neurodegeneration.

The purpose of this study is therefore to assess, in a large cohort cognitively normal subjects characterized by APOE genotype, cross-sectional associations between multimodally-measured CP structure (MRI-measured overall CP volume and CT-measured CP calcium volume) within the lateral ventricles, PET-measured Aβ burden and neurodegeneration. We use MRI-measured hippocampal volume as a proxy for Alzheimer’s disease–related neurodegeneration, consistent with multiple studies showing that hippocampal atrophy is a robust and well-validated biomarker of neurodegeneration in AD33,34. We quantify the relative contribution of APOE genotype, CP calcium volume, CP overall volume and Aβ burden to hippocampal volume. The long-term objective of this line of research is to determine whether CP structural abnormalities, including CP calcification measurable using widely-available CT, can serve as a biomarker of AD risk that can guide future targeted AD prevention strategies.

Methods

Subjects and assessment

Subjects included all cognitively normal subjects who underwent PiB PET-CT and MRI between 2009 and 2022 as part of several studies of aging conducted at Weill Cornell Medicine and New York University. Subjects were recruited from the community via advertising. Each study had slightly different inclusion criteria, but all required subjects to be medically healthy apart from common condition such as hypertension or hypercholesterolemia, without significant psychiatric disease and free from alcohol or substance abuse. While some studies enrolled patients with a known or suspected diagnosis of MCI or AD, this re-analysis was limited to subjects with normal cognition (defined below.) All studies were approved by the institutional review boards and have been performed in accordance with the Declaration of Helsinki. All subjects formally consented to participate in the study by completing and signing the required written consent forms.

Subjects underwent detailed assessment using the National Alzheimer’s Coordinating Center (NACC) assessment battery35 including the Clinical Dementia Rating (CDR) scale administered by a specialist physician. Assessment typically occurred within 6 months of neuroimaging. Subjects included in this analysis had CDR = 0, indicating no meaningful cognitive impairment.

Image acquisition

Image acquisition for this study took place over 15 years and two PET and three MRI scanners. Key acquisition parameters are summarized below with details provided in Supplementary Materials (Tables S1-S2). Potential confounding effects of different scanners and acquisition parameters were included as covariates in analysis.

PET acquisition and processing

PET scanning was performed at Weill Cornell Citigroup Biomedical Imaging Center on a General Electric (GE) Discovery LS PET/CT from 2009 to 2011 (G.E. Medical Systems, Milwaukee, WI) then on a Siemens Biograph PET/CT from 2011 to 2022 (Siemens, Knoxville, TN). Subjects were injected with ∼550 MBq of locally synthesized 11 C-PiB. Scanning acquisition differed slightly between 2009 and 2022 but always included the 60–90 min period post-injection, which was analyzed for this study. PET voxel size ranged from 1.9 × 1.9 × 4.3 mm in the earliest scans to 1.0 × 1.0 × 2.0 mm more recently. Additional acquisition parameters are provided in Supplementary Materials (Table S1).

PET images were corrected for photon attenuation, scatter, and radioactive decay and processed using an in-house image processing pipeline validated and used previously36. Briefly, processing consists of aligning dynamic PET frames to the first frame using rigid-body registration and averaging them to generate a static PET image, which is then co-registered to that subject’s T1 MRI. Average PiB SUVr (reference = cerebellar gray) is then calculated within a Freesurfer-defined meta-region consisting of AD-vulnerable cortical regions (cingulate, precuneus, frontal, parietal and temporal lobes)37.

CT acquisition

Before each PET acquisition, a CT scan was acquired for attenuation correction and neuroanatomic localization. CT acquisition parameters varied significantly over time due to hardware changes and increased focus on minimizing radiation exposure. On the GE scanner (2009–2011) tube current was 125 mA and current-exposure time product (exposure) was 1104 mAs. On the Siemens scanner (2011-present) automatic exposure control was instituted to reduce patient radiation exposure; tube currents ranged from 22 to 216 mA and exposure from 14 to 180 mAs. Visual review of CT scans showed decreased image quality (increased noise/graininess) for scans acquired after automatic exposure control was instituted. CT acquisition parameters (tube current) were therefore included as a covariate in statistical analysis.

MRI acquisition and processing

MRI was acquired on either a 1.5T GE Signa scanner or one of two 3T Siemens Prisma scanners. T1w and FLAIR images were used in analysis. Acquisition parameters, which varied over time and scanners, are provided in supplemental data. Notably, the earliest FLAIR images included as few as 32 3.3 mm slices; these were manually reviewed to ensure they included the full CP. Additional acquisition parameters are provided in Supplementary Materials (Table S2).

T1 MRI

T1 MRI was used for PET and CT co-registration and regions of interest (ROI) segmentation. T1 MRI images were processed with FreeSurfer version 7 (http://surfer.nmr.mgh.harvard.edu) recon-all pipeline with manual editing of the white and gray matter borders to correct any mis-segmentation. The following bilateral ROIs were used in analysis: hippocampi, cerebellar gray matter, a cortical meta-region consisting of AD-vulnerable regions37, lateral ventricle and CP. The hippocampal ROI was divided by total intracranial volume (TICV) calculated using SPM12 (https://www.fil.ion.ucl.ac.uk/spm) and used as an index of neurodegeneration. Cerebellar gray matter was used as the reference for PiB PET SUVr calculation. Lateral ventricle (all horns: frontal, temporal and occipital) and the CP within the ventricles were combined into a ventricle + CP ROI. This combined ventricle + CP ROI identified the region where CP could be located, and was used in the algorithms described below to accurately measure CP overall volume and CP calcium volume through co-registration to different image modalities (FLAIR MRI for CP overall volume; CT for CP calcium volume.) This was necessary because visual review of the CP segmentation based on T1 MRI showed significant inaccuracies and calcium is not visible on T1 MRI.

FLAIR MRI

Volume of CP within all horns (frontal, temporal, occipital) of bilateral lateral ventricles was measured by co-registering T2-FLAIR images to the T1w anatomical space used for FreeSurfer segmentation. The ventricle + CP ROI defined on T1 MRI (as described above) was eroded by one voxel to avoid partial volume effects. The FLAIR image in the eroded LV + CP mask was then denoised using a 3D Gaussian filter with sigma = 0.5 and fitted into a Gaussian mixture model (GMM) with three components. Voxels that belonged to the cluster with the highest average intensity value were selected as the final CP segmentation. This method was validated by comparing it to expert manual tracing using an independent dataset of 40 subjects, showing an average Dice similarity coefficient of 0.82, indicating high spatial overlap, reproducibility and accuracy38.

CP calcium quantification

CP calcium was measured using a validated automated method described previously20. Briefly, CT was co-registered to T1w MRI and CT voxels within the Freesurfer-defined LV + CP mask with Hounsfield units (HU) greater than 80 were used as seeds to segment contiguous areas of calcium with the ventricle. We previously showed that this method had intraclass correlation coefficient of 0.982 with expert visual tracing20.

All segmentations were reviewed manually for accuracy. As noted above, CT images acquired after 2011, when automatic radiation exposure control was instituted on the Siemens PET-CT scanner, had noticeably reduced quality, though our CP calcium segmentation algorithm remained accurate. For this reason, tube current was included as a covariate in statistical analyses. CP calcium volume was normalized by TICV.

APOE genotype determination

APOE4 status (presence or absence of an APOE4 allele) was determined using standard polymerase chain reaction procedures. Study subjects were classified as APOE4 carriers (APOE4+) if they had at least one APOE4 allele and classified as APOE4 non-carriers (APOE4−) otherwise.

Statistical analysis

Analyses were performed in R and SPSS.

Descriptive statistics

Continuous variables (age, hippocampal volume, CP calcium, CP overall volume, PiB SUVr) were checked for normality using Shapiro-Wilk test and summarized by mean/std or median/range for the full group, and separately in APOE4 carriers and non-carriers.

Group comparisons between APOE4 carriers and non-carriers

Comparisons of continuous variables between APOE4 carriers and non-carriers used t-test (for normally distributed variables) or Mann Whitney test (for non-normally distributed variables). Comparisons of categorical variables (sex, binarized presence vs. absence of CP calcium) between APOE4 carriers and non-carriers used chi-square test.

Correlations

We used Spearman correlation to assess pairwise associations between continuous variables in the full sample and separately in APOE4 carriers and non-carriers. Because all variables were expected to correlate with age, analyses were conducted both with and without adjusting for age (i.e., using partial Spearman correlations). To control for multiple comparisons, p-values were corrected using the false discovery rate (FDR) method.

Multiple linear regression

All regression assumptions (normality, linearity, homoscedasticity, absence of multicollinearity) were checked. We used a variance inflation factor (VIF) < 5 as the threshold for multicollinearity39. Boxplots were used to identify potential outliers.

We used multiple linear regression with dependent variable = hippocampal volume (our index of neurodegeneration) and key predictors = APOE4 genotype, CP calcium, CP overall volume, cortical PiB SUVr, age and sex. We included the following potentially confounding neuroimage acquisition parameters: MRI field strength (3T versus 1.5T), days between MRI and PET-CT scanning and CT tube current, which varied significantly over time due to institution of automatic radiation exposure control in 2011, and which affects CT image quality/noise and thus CP calcium quantification. Because prior studies have implicated APOE4 status as highly relevant to AD risk4,5,6, and possibly to CP structural changes24, we assessed potential interactions between APOE4 status and CP measures.

Results

Subjects

From a total of 206 PiB PET scans from cognitively normal subjects, 40 were excluded for the following reasons: 36 subjects lacked either T1 or FLAIR MRI within 1 year of PET, 2 had FLAIR scans that were not full-brain and did not include CP, 1 PET scan had abnormally low counts, 1 subject with hippocampal volume 2 std below mean was re-reviewed and re-classified as cognitively impaired (MOCA = 16). This left 166 subjects. 21 subjects (12.7%) were Aβ + based on PiB SUVr > 1.3 in cortical region confirmed by visual interpretation.

Demographics and other characteristics of subjects are presented in Table 1.

Table 1 Participant characteristics and neuroimaging parameters.
Full size table

Group differences between APOE4 carriers and non-carriers

As shown in Table 1, APOE4 carriers and non-carriers did not differ in terms of demographics, CP overall volume, CP calcium, or any other measured variables. 9/63 (14.3%) of APOE4 carriers and 15/103 (14.6%) of APOE4 non-carriers. had no detectable CP calcium; these percentages did not differ significantly (p = .99). Results were unchanged when one APOE4 + subject with the largest CP volume was excluded from analysis.

Correlation results

All correlation results are presented in Supplementary Materials.

We examined pairwise Spearman correlations among all continuous variables: age, CP overall volume, CP calcium volume, PiB SUVR, and hippocampal volume in the full sample (N = 166), as well as in APOE4 carriers and non-carriers separately. In the unadjusted full-sample analysis, age was significantly associated with greater CP volume (ρ = 0.403, p < .001), higher PiB SUVR (ρ = 0.526, p < .001), and lower hippocampal volume (ρ = –0.506, p < .001). CP overall volume correlated positively with PiB SUVR (ρ = 0.225, p = .004) and negatively with hippocampal volume (ρ = –0.374, p < .001). CP calcium volume was significantly associated only with CP overall volume (ρ = 0.327, p < .001). These patterns were broadly similar when stratified by APOE4 status.

After controlling for age, most associations were attenuated. In the full group, only the correlations between CP overall volume and CP calcium volume (ρ = 0.308, p < .001), and between CP overall volume and hippocampal volume (ρ = –0.216, p = .005), remained significant. In the age-adjusted, APOE-stratified analyses, a genotype-dependent effect emerged: the correlation between CP overall volume and hippocampal volume was significant in APOE4 carriers (ρ = –0.420, p < .001), but not in non-carriers (ρ = –0.081, p = .417), and this difference was statistically significant (p = .025, two-tailed Fisher r-to-z test). This suggests that, independent of age, CP enlargement is associated with reduced hippocampal volume only in individuals carrying an APOE4 allele.

Multiple regression results

One subject with the largest CP overall volume (6.52 vs. median = 1.32), who was APOE4-positive, was identified as an outlier based on boxplot analysis and was excluded from analysis. As shown in Table 2, age, sex, CP overall volume, and the interaction between APOE4 status and CP calcium were found to be significant predictors of hippocampal volume. Women had larger hippocampi (normalized by TICV) than men. The interaction between APOE4 status and CP calcium was due to a significant difference in the relation between CP calcium and hippocampal volume in APOE4 carriers as compared to non-carriers, as shown in Fig. 1. No other interactions were significant.

Table 2 Multiple regression model (R2 = 0.345, R2 adjusted = 0.30, p < .001) predicting hippocampal volume in 165 cognitively normal subjects.
Full size table
Fig. 1
figure 1

Divergent association between choroid plexus calcium and hippocampal volume in APOE4 carriers vs. non-carriers. Scatter plot showing the relationship between CP calcium volume and hippocampal volume, both normalized by total intracranial volume, stratified by APOE4 status. The differing slopes illustrate the significant interaction between CP calcium and APOE4 status observed in the regression analysis, after adjusting for additional variables.

Full size image

To clarify the basis for this interaction, we performed multiple regression analysis for APOE4 carriers and non-carriers separately. We found that in APOE4 carriers, CP overall volume (β=-0.255, t= -2.481, p = .016), sex (β=-0.338, t=-2.190, p = .033), and at a trend level, CP calcium (β = 0.639, t = 1.900, p = .063), age (β=-0.013, t=-1.862, p = .068), and the CT tube current (β = 0.003, t = 1.809, p = .076) predicted hippocampal volume. In APOE4 non-carriers, only age (β=-0.018, t=-4.471, p < .001) predicted hippocampal volume. The inverse association of CP overall volume and hippocampal volume (greater CP volume associated with reduced hippocampal volume) only in APOE4 carriers in this analysis mirrors results from the correlation analysis.

Discussion

This first neuroimaging study linking multimodally measured CP volume and calcification to AD neuroimage biomarkers contributes to the growing body of research linking CP structural abnormalities to neurodegeneration and AD pathogenesis, highlighting that the important role of CP in maintaining brain health and homeostasis extends far beyond its traditionally-recognized function of CSF production. We found, in 166 well-characterized, cognitively normal subjects, that MRI-measured CP enlargement was associated with reduced hippocampal volume – an index of neurodegeneration, with this association strongest in APOE4 carriers. This results adds to mounting evidence that CP enlargement is pathological24,25,26,27,28. CP calcium was not independently associated with hippocampal volume. However, a significant interaction revealed APOE4 genotype-specific associations between CP calcium and neurodegeneration. As discussed below, we believe these findings offer important mechanistic insight into how choroid plexus structural alterations may interact with APOE4 genotype to influence vulnerability to neurodegeneration and drive key processes in AD pathogenesis.

The now well-established observation that CP enlargement signifies pathology, in contrast to the rest of the brain where pathology is marked by volume loss / atrophy, prompts the critical question: what is the basis of CP enlargement? To investigate this, we used multimodal neuroimaging to measure the volume of calcium deposited within CP. We used the CT scan acquired at the time of PET (for attenuation correction) and a validated CT-MRI segmentation method we developed20 to quantify calcium, which cannot easily be quantified accurately using MRI40,41. Based on clinical knowledge that CP calcification increases with normal aging and in certain forms of neurodegeneration29,30,31,32, and our recent finding that cortical inflammation correlated closely with CP calcium volume (more so than CP overall volume)20, we had hypothesized that the volume of calcium within CP might be a more accurate/sensitive biomarker of CP pathology than CP overall volume, and might therefore be more closely linked to neurodegeneration. This hypothesis was not confirmed. Even though the volume of calcium within CP correlated with overall CP volume (as expected, and as we showed previously20), we found no evidence that the volume of calcium within CP was more specifically associated with neurodegeneration. In fact, we found an unexpected trend towards the opposite in APOE4 carriers. As detailed below, we believe these intriguing dissociations in the correlation of CP calcium as compared to CP overall volume with hippocampal volume, and in APOE4 carriers versus non-carrier, provide novel insight into the role of CP in neurodegeneration, and how the APOE4 genotype promotes AD pathogenesis.

Greater MRI-measured CP volume is associated with reduced hippocampal volume – an index of neurodegeneration

Using simple correlation analysis (with or without controlling for age) as well multiple regression that accounted for additional relevant biological factors (age, sex, APOE4 status) and potential methodologic confounds such as neuroimage acquisition parameters, we found that larger MRI-measured CP overall volume was associated with lower hippocampal volume – an index of neurodegeneration. Larger CP overall volume was also associated with older age, though the association of CP overall volume and hippocampal volume was independent of age. These findings align with prior MRI studies of CP structure and confirm that increased CP volume is pathological24,25,26,27,28.

The association between larger overall CP volume and reduced hippocampal volume appears to be driven by APOE4 carriers

When correlation and regression analyses were performed separately for APOE4 carriers and non-carriers, a significant inverse association between CP volume and hippocampal volume was observed only in APOE4 carriers. Although the interaction effect (genotype by overall CP volume) was not significant in the full group analysis, results indicate that the inverse relationship between CP overall volume and hippocampal volume may be more pronounced in APOE4 carriers. These findings are broadly consistent with a prior study demonstrating that APOE4 carrier status is associated with markedly accelerated choroid plexus enlargement over time, suggesting that the E4 allele may influence dynamic CP changes during aging24.

Significant interaction between APOE4 genotype and CP calcium on hippocampal volume

Multiple regression analysis did identify a significant interaction between APOE4 genotype and CP calcium volume on hippocampal volume (p = .047). As illustrated in Fig. 1, in APOE4 carriers, there was a trend (p = .063) suggesting that higher CP calcium was associated with greater hippocampal volume. In contrast, APOE4-negative individuals showed an opposite pattern.

The pattern in APOE4 non-carriers is expected. Decades of clinical and research work, starting with pneumoencephalography in the early 1900s, has documented CP calcification in association with normal aging and in some forms of neurodegeneration29,30,31,32. Based on this, we had hypothesized that CP calcium volume would be inversely associated with an index of neurodegeneration (hippocampal volume). However,

this hypothesis was not confirmed, and in APOE4-positive subjects, we found trend-level evidence for the opposite.

The association of neurodegeneration with greater CP overall volume but (in APOE4 carriers) lower CP calcium indicates that a substance/process other than calcium underlies pathologic CP enlargement, and this substance/process may be especially important in E4 carriers

The unexpected positive, trend-level association of greater CP calcium volume with greater hippocampal volume in APOE4 carriers, coupled with a strong, inverse association between CP overall volume and hippocampal volume, indicates that the deposition of substances other than calcium within CP and/or other structural CP alterations are responsible for the link between CP enlargement and neurodegeneration. Understanding what other than calcium could be causing pathologic CP enlargement, which we and one prior study24 have shown to be most prominent in E4 carriers, has the potential to provide insight into the broader mechanisms by which the APOE4 allele promotes the AD pathogenetic process. Mechanisms proposed to underly the AD-promoting effect of the APOE4 allele include more severe blood-brain-barrier breakdown, greater inflammation, abnormal iron metabolism and lipid dyshomeostasis5,6,7. Any of these mechanisms could be relevant to our results given that that inflammatory cells, iron and cholesterol are all known to accumulate in CP19,25,42. Particularly intriguing is a recent human neuropathology demonstration of greater lipid deposition and associated complement-mediated inflammation in CP from E4 carriers with or without AD43. That study also found that the amount of lipid deposited within CP correlated with the extent of Aβ and tau deposition43. An in vitro study found that APOE4 is a less effective acceptor of cholesterol excreted from CP epithelial cells, causing cholesterol accumulation within CP in E4 cells44. Our in vivo results, coupled with these neuropathological and mechanistic studies, implicate CP as a key locus for understanding how the APOE4 allele promotes the AD pathogenic process.

Sex differences

Women had larger hippocampi (normalized by TICV) than men. This has been reported previously45,46, though the magnitude of this effect in our small sample is somewhat surprising. The female sex hormone estrogen has important, beneficial effects on hippocampal structure47,48, and decreased estrogen along with compensatory increased gonadotropins at menopause are considered a key explanation for increased rates of AD in women after menopause49,50. It is a limitation of this study that we do not have information about subjects’ menstrual or menopausal status in order to better interpret this finding. We controlled for sex in all analyses.

Lack of association of CP structure with Aβ deposition

We were not able to demonstrate an association between CP structure (neither overall nor calcium volume) and level of Aβ deposition. The association between CP overall volume and Aβ in our correlation analysis was caused by age, since both increase with age. Given that preclinical and neuropathological studies implicate CP failure in Aβ accumulation11,12,25,43, with mounting evidence that CP dysfunction may be reflected in MRI-measurable enlargement, it would be premature to give up on demonstrating an association between CP structure and Aβ deposition. A key limitation of our analysis is that we did not incorporate Bayes factors, which means we cannot support the absence of a relationship between CP structure Aβ; rather, we can only conclude that we failed to reject the null hypothesis of no relationship51. Our negative findings may relate to limited sample size and our inclusion of only cognitively normal subjects with a small number of Aβ + subjects (21/166; 12.7%). It is also possible that CP structural alterations are more associated with tau than Aβ52,53. Future studies should assess CP structure in association with wider range in level of Aβ deposition, as well as in association with brain tau deposition.

Additional limitations

Because scans for this study were acquired over many years, there were changes in scanner hardware, which affected some measures. We used the CT acquired with PET-CT to measure CP calcium. Quantifying brain calcium is challenging using MRI, and CT is considered the gold standard for this. However, over recent years, in order to minimize unnecessary radiation exposure, the radiation dose of CT needed only for PET attenuation correction has been significantly reduced, and automated exposure control implemented. Addressing this limitation, we included CT parameters, as well as MRI parameters (3T vs. 1.5T) as covariates in analysis. In the near future, calcium sensitive MRI methods54 may obviate the need for CT to measure brain calcification.

Another limitation of our study is the reliance on hippocampal volume as the sole measure of neurodegeneration. While MRI-measured hippocampal volume has been shown to correlate with neuropathological evidence of neuronal loss and AD pathology34, it is not a direct marker of neurodegeneration. Instead, progressive hippocampal atrophy over time provides a more accurate reflection of the neurodegenerative process. The absence of longitudinal data in our study is therefore a significant limitation, as it prevents us from assessing changes in hippocampal volume, as well as meaningful change in cognition.

Conclusion

Our results demonstrate that MRI-measured CP enlargement is associated with neurodegeneration, indexed as lower hippocampal volume, in cognitively normal subjects, with this effect driven by APOE4 carriers. We also confirmed greater CP volume in association with older age. These findings add to the growing body of research indicating that CP enlargement is pathological24,25,26,27,28, and emphasize that the important role of CP in maintaining brain health and homeostasis extends far beyond its traditionally-recognized function of CSF production.

With the goal of understanding the basis for pathologic CP enlargement, we used a validated, multimodal CT + MRI technique to measure the volume of calcium deposited within CP20. We had thought CP calcium would be a more sensitive measure of CP pathology than overall CP volume, but this hypotheses was not confirmed. What we did find is a divergent association between CP calcium and hippocampal volume depending on APOE genotype. Specifically, we found a trend towards greater hippocampal volume with greater CP calcium volume in E4 carriers – the opposite of what we had hypothesized. This suggests that pathologic CP enlargement in association with neurodegeneration is due to substances and/or processes other than calcium, with these substances/processes especially relevant in E4 carriers.

Based on recent neuropathology study43, a likely substrate for pathological, MRI-measurable CP enlargement that is more prominent in E4 carriers, may be lipids and inflammatory cells. This study demonstrated greater lipid and associated complement-mediated inflammation in the CP epithelia of E4 carriers, and in correlation with AD lesions43. Additional work, including direct correlation between in vivo neuroimaging findings and ex vivo neuropathology would be needed to confirm this. Investigating APOE genotype-specific changes in CP structure can provide valuable insight into the broader mechanisms by which the APOE4 allele contributes to AD pathogenesis, which remain highly debated5,6. The present neuroimaging findings in combination with available neuropathological studies43 offer support for dysregulated inflammation and lipid metabolism as key to understanding of how AD progression may differ between APOE4-positive and APOE4 − negative individuals5,6,55, and highlight the CP as a key locus of these differences. Such insights are of considerable interest and therapeutic relevance, particularly given the recent availability of monoclonal antibody therapies for AD, which have demonstrated an unfavorable risk/benefit profile in APOE4 homozygotes56.