Cerebrovascular Disease Underlying White Matter Pathology in Alzheimer’s Disease with and without a History of COVID-19

Principal Investigator: Faye McKenna, PhD

The McKenna Lab and research team at the Magnetic Resonance Research Center

1. Overview

A common neurological finding in Alzheimer’s Disease (AD) and those affected by Covid-19 are increased white matter hyper-intensities (WMH), which are accompanied by microvascular pathologies and demyelination^2-3. Our study aims to investigate two cohorts of AD (N = 240) and AD with a history of COVID-19 (N = 200) to investigate how cerebrovascular pathology contributes to WMHs and clinical manifestations based on advanced neuroimaging methods. We believe new therapies could be designed to treat cerebrovascular biotypes. This funding will support an early-stage investigator’s project, specifically postdoc assistance.

2. Introduction

BACKGROUND AND SIGNIFICANCE: Alzheimer’s disease (AD) is a devastating neurodegenerative illness which represents one of the leading causes of disability worldwide, affecting approximately 10% of those 65 years and older, ~55 million people worldwide¹. Those affected with AD have some of the highest rates of early death and in addition to causing family and caregiver distress, AD is estimated to accounting for 2.5% of yearly United States expenditures spent on mental illness¹. A major contributing factor to the disorders’ low treatment response and burden is our lack of understanding of dementia’s underlying pathophysiology, which is commonly associated with two defining pathological features-the accumulation of plaques and tangles, demyelination/atrophy and cerebrovascular dysfunction most clearly seen as white matter hyper-intensities (WMH)². Since the COVID-19 pandemic, there has been an uptick in both AD and cerebrovascular disorders with research suggesting COVID-19 can contribute 50-80% to the risk of developing AD³.

Accumulating evidence from histological, immunochemical, genetic, and neuroimaging studies implicates cerebrovascular pathology in AD^4–7. Additionally, AD is more prevalent in those with existing cardiovascular risk factors^8,9. Recent drug trials have demonstrated that drugs with cerebrovascular mechanisms, i.e. blood brain barrier (BBB) repair, blood pressure regulation, increasing cerebral blood flow, can alleviate symptoms in a portion of patients with AD¹⁰. Since cerebrovascular dysfunction has been shown to vary across AD¹¹, these drug trials would substantially benefit from using AD cerebrovascular biotypes to provide critical information as to which patients may be helped by vascular treatments.

A major barrier to establishing cerebrovascular biotypes has been a lack of suitable biomarkers. intravoxel incoherent motion imaging (IVIM) is a simple diffusion weighted (DWI) MRI method that was originally proposed in the 1980’s and measures capillary perfusion via the diffusion-like signal arising from blood protons flowing through randomly oriented tissue capillaries¹². Despite IVIM being a long-standing and well-tested method, the IVIM method has mostly been used to measure microvascular perfusion in the liver and kidney and its application to the brain has not been very well explored¹³. This is likely due to the popularity of other perfusion quantification methods like arterial spin labelling (ASL) and dynamic contrast (DCE/DSC) imaging. However, recent research shows IVIM may be superior to these other perfusion MRI techniques in quantifying small capillary pathology rather than larger arteriole changes^14–17. The IVIM technique is also highly clinically translatable, only requiring a short ~1.5-minute scan which can also be combined with standard DWI imaging for more advanced modeling and does not use contrast. No research has been conducted to leverage the IVIM method to delineate cerebrovascular biotypes in AD, particularly those with AD who have been affected by COVID-19.

In recent years the free-water imaging (FWI) technique has gained popularity for its ability to measure inflammation in the brain, with the free water (FW) metric which has been shown to associate with activated microglia and astroglia, and most recently a leaky blood-brain barrier¹⁸. Rydhog and colleagues expanded upon this model to create the IVIM-FWI model which uses a multi-shell acquisition and includes lower b-values to integrate the IVIM pseudo-diffusion microvascular high-diffusivity compartment¹⁹. The fraction of the microvascular compartment derived using IVIM-FWI, the perfusion fraction (PF), has been shown to reflect capillary density and vasodilatation¹⁴ and the combined FWI-IVIM model was shown to substantially improve the FW estimation in addition to estimating PF.

Numerous studies conducted in rodent disease models have found that altered IVIM-PF metrics associate with microvascular pathologies in the brain such as vascular density²⁰, microvascular circulation²¹. Notably for AD, increased IVIM PF has been correlated with white matter hyperintensities, a hallmark sign of cerebrovascular pathology a strong correlate of neurocognitive functioning^22,23. Additionally, our preliminary work demonstrates that PF is significantly decreased in those with AD and AD that have also had at least one episode of COVID-19 and better associate with WMH compared to the ASL-derived cerebral blood flow (CBF) metric (Figures 1&2).

Figure 1. Decreased IVIM PF in 1) an AD patient, and 2) an AD patient with a history of COVID-19. A-C) shows the FLAIR WMH which were segmented in red, D-F) shows the IVIM PF with the WMH outline in yellow, and G-H) shows the ASL CBF with the WMH outline in yellow with less apparent overlap.

Figure 2. A) Freesurfer segmentation of WMH in blue, B) correlation between IVIM PF and ASL CBF perfusion measures in the three tissue types, C) FLAIR intensity of three tissue types, D) IVIM PF intensity of the three tissue types with separation between normal and abnormal WMH, and E) ASL CBF intensity of the three tissue types with no separation between normal and abnormal WMH.

We propose that (1) used together in a multi-modal fashion, IVIM-FWI-derived PF and FW and ASL-derived CBF will be able to quantify anatomically distinct microvascular properties of cerebrovascular pathology in those with AD and AD with COVID-19 and explain the underlying pathology behind WMH/WM pathologies; (2) be used non-invasively, including radiation and contrast-free, in a clinical setting that has generalizability; and (3) have whole-brain high-resolution coverage.

We will test if PF, FW and CBF measures in two data sets of patients with AD and AD with a history of COVID-19: (Aim 1) are sensitive to cerebrovascular pathology in patients compared to controls; (Aim 2) able to delineate two cerebrovascular biotypes within the patient groups; and (Aim 3) uniquely associated with cognitive impairment and peripheral biomarkers of vascular health.

INNOVATION: The novelty of this study comes from: 1) Using a muli-modal in vivo approach which combines IVIM-FWI multi-shell DWI alongside ASL that is able to quantify different aspects of cerebrovascular pathology in the whole brain and WMH and tests IVIM’s unique clinical translatability as a tool for measuring perfusion in AD. 2) Examining novel samples of those with AD and AD with a history of COVID-19 with the aim of finding biotypes based on neurobiological cerebrovascular markers with vascular treatment target potential. The recent research suggesting COVID-19 may increase the risk of developing AD up to 80%³ makes it an unprecedented opportunity to isolate microvascular biotypes in AD with treatment potential.

3. Methods

APPROACH: This proposed analysis will be conducted on two data sets: 1) existing data of N = 240 participants; 160 AD patients along with 80 matched healthy controls taken from the Einstein Aging Study (EAS), and 2) newly acquired data N = 200 participants; 120 AD patients with COVID-19 along with 60 matched healthy controls from a U19 study on Drug Discovery for Aging (U19); with the aim of completing enrollment by August 2026. All participants are evaluated using several cognitive tests under the supervision of a licensed clinical psychologist.

All MRI data are acquired on a 3T Prisma Siemens MRI using a body coil for transmission and a 32-element coil for reception. In addition to the diffusion data, T1, FLAIR and T2-weighted images are obtained for clinical evaluation and high-resolution atlas registration in each subject. Anatomical data are acquired using a voxel size of 1x1x1 mm³⁷³. Multi-shell EAS Diffusion MRI data were acquired with b-values of 0, 300, 800 and 2000 s/mm² and the U19 Diffusion MRI data are acquired using 0, 500, 1000 and 2000 s/mm² b-values, all with a 1.5x1.5x1.5 mm³ isotropic voxel size. This grant funding will allow us to add the additional 1.5 minutes of scanning to the U19 data protocol to obtain the additional 50, 100 and 250 s/mm² b-values to measure the IVIM effect (which can be detected only under b = 500 s/mm²). IVIM-FWI images will be processed using a combination of mrtrix²⁴, FSL²⁵, FreeSurfer²⁶ and DIPY²⁷ software and data pre-processing will follow a standard processing pipeline and include denoising and Gibbs artifact removal, image quality check, and motion and image distortion correction. IVIM-FWI metrics will be calculated using the multi-shell estimation technique as previously described by Rydhog et al. using DIPY software^19,27. The EAS ASL data was acquired with a single delay pseudo-continuous arterial spin labeling (pCASL) pulse and the U19 ASL pCASL pulse, both 2.5X2.5X4.5 mm³ voxel size and pre-processing and CBF calculations will be carried out using FSL’s BASIL package²⁵.

The Human Connectome Project (HCP) structural processing pipelines for surface generation and alignment to the Desikan-Killiany atlas will be used on T1 and T2 weighted images. Subsequently, each subject-specific cortically-labelled volume and surface obtained from the structural processing will be warped into the DWI or ASL space using HCP processing pipeline²⁸. Mean PF, FW and CBF values will be calculated across gray and white matter voxels, within the Desikan-Killiany atlas ROIs and WMH segmentation and additionally projected onto individual subjects’ cortical surfaces to assess metrics in vertex-wise manner using the HCP’s surface registration pipeline.

SPECIFIC EXPERIMENTS: Aim 1: will test for between-group differences in several imaging markers known to be associated with vascular processes in the brain. These markers include PF, FW and CBF. The analyses will test differences in several brain regions found to be most affected by AD pathology. These regions include the temporal lobe and its 8 sub-divisions, frontal lobe and its 11 subdivisions, amygdala, thalamus, cingulate, superior parietal lobe, precuneus and WMHs as delineated by the Desikan-Killiany atlas provided with FreeSurfer software²⁶.

The following approach will be undertaken. First, independent samples T-tests will be applied to compare AD patients to healthy controls (HC) in terms of each individual regional imaging measure. The sample sizes of a) 160 AD patients and 80 healthy controls from the previously acquired EAS sample, and b) 50 AD patients with COVID-19 and 50 HC from the U19 sample are deemed sufficient for an asymptotic justification of the use of these parametric tests. Second, logistic regression will be performed to assess the ability of PF, FW and CBF metrics, first individually and then together, in the a priori ROIs to discriminate between AD and healthy control participants in each sample. Specifically, binary logistic regression with elastic net regularization (ENR) will be used to identify a set of features representing significant independent predictors of group membership and to fit a model to predict AD as a function of the selected features. ENR is a shrinkage operator which includes LASSO and ridge regression as special cases and performs automated variable selection while mitigating problems associated with multicollinearity and overcoming some of the limitations associated with LASSO. In addition to the imaging measures, the matching factors (age, sex, socioeconomic background and premorbid IQ) will be allowed to compete for inclusion in the prediction model. These tests should provide both a detailed in vivo map of gray matter and WMH areas with microvascular differences in AD/AD with a history of COVID-19 and highlight the potential clinical diagnostic value of the proposed IVIM-PF metrics in diagnosing AD.

Anticipated results and alternative outcomes: The T-tests are expected to detect significantly decreased PF metrics of microvascular perfusion in WMHs, like what we detected in our preliminary data analyses in additional to the a priori ROIs deemed to be most related to AD (Figures 1&2). We expect to see increased FW and decreased CBF, but for PF to be the strongest marker separating groups. Although we believe our preliminary results and others work are a strong indicator that the proposed study will identify group difference in both samples, we may fail to find group differences due to higher-than-expected variability in the levels of microvascular abnormalities within AD and AD with a history of COVID-19. Therefore, to further explore the regional differences in perfusion we will use the HCP pipeline multi-modal surface matching and voxel-based comparison tools, which include corrections for multiple comparison, to help identify other brain regions showing significant differences in terms of the PF, FW and CBF measures between groups²⁸.

Aim 2: will examine if in vivo MRI perfusion measures can delineate two cerebrovascular biotypes within AD. It is clear that cerebrovascular heterogeneity exists across and within AD, with previous research pointing to there being about 24% with no vascular pathology and 50% with vascular lesions²⁹. We will perform agglomerative hierarchical clustering using the cerebrovascular measures (PF, FW and CBF) in the ROIs as the input to partition the sample of AD patients into clusters that are internally homogenous (minimal within subgroup variation) and externally heterogeneous (substantial variation between subgroups) in terms of the imaging markers of cerebrovascular pathology. Various partitions will be considered by setting the number of clusters at 2, as is consistent with prior studies and our stated objective, and by setting the number of clusters at values identified as optimal according to each of the Silhouette and Dunn cluster validation indices. This analysis will investigate heterogeneity and patterns of in vivo cerebrovascular metrics in AD in order to isolate biotypes that may be suitable for promising new vascular-based therapies¹². The clinical validity of the resultant clusters as biotypes will be assessed in Aim 3, wherein the clusters will be compared in terms of clinical features and blood biomarkers of inflammation.

Aim 3: will test whether AD cerebrovascular biotypes are associated with clinical features and peripheral vascular markers. We will use several analysis of covariance (ANCOVA) tests to compare the resultant cerebrovascular biotypes within each set of clusters from Aim 2 in terms of memory and AD severity measures in order to provide information as to how the quantified cerebrovascular biotypes present as AD phenotypes. Secondly, we will use ANCOVA tests, with age, medication, pre-morbid IQ and sex as covariates, to compare Aim 2’s cerebrovascular biotypes in terms of blood pressure, heart rate and vascular disease. Based on previous research, we hypothesize that the biotypes containing patients with greater cerebrovascular pathology metrics will associate with greater illness duration, age, increased symptoms and peripheral vascular pathology⁸. Relating external measures of peripheral vascular pathology and clinical presentation to the cerebrovascular biotypes should support the distinctiveness and value of the proposed metrics in detecting cerebrovascular pathology.

POWER ANALYSIS: Preliminary data for PF within 54 brain regions showed the metric had a coefficient of variation (CV) < 10% within each subject group for more than 50% of the brain regions that were considered. For the majority of regional imaging measures that are expected to have a CV < 10% within each group, the Aim 1 T-tests that compare the patients and controls with a familywise type I error rate of 5% after a Benjamini-Hochberg multiple comparison correction, will have 80% power to detect a mean difference of 10% between AD patients and controls when the difference is expressed as a percentage of the mean for controls. The power of the Aim 3 tests to compare the biotypes in terms of clinical outcomes will depend on the unknown number of AD patients assigned to each biotype with power being highest when the biotypes have the same sample size. Since the identification of biotypes will have limited clinical utility if one of the biotypes is relatively uncommon, power was assessed for the statistically optimal case where 50% of the patients have a given biotype and for the conservative yet still clinically relevant case where one biotype consists of only 25 of the 75 patients. Conservatively assuming the outcomes measures to have a CV < 25% within each biotype, there will be 80% power at the two-sided 5% significance level to detect a mean difference of 18% for the optimal case and a mean difference of 20% for the conservative case.

Facilities and Ethics Approvals: The project requests $5,000 USD for essential personnel. Our MRI research center at the Albert Einstein College of Medicine are well-equipped to perform the proposed research with high-powered computing clusters, and the principal investigator is currently approved for the proposed experimental procedures working with human data and to have this contribute to mentoring a postdoc student for the year.

Estimated Timeline: This project is designed for completion within 12 months.

Data Sharing: We are committed to open science and will share data generated in this project through an open-access public repository (e.g., OSF, GitHub) upon publication.

4. Pilot Data (Optional)

See above Figures demonstrating the ability of IVIM to delineate WMH microvascular pathology in two individuals with AD and AD and a history of COVID-19.

5. Budget

This research proposal has already collect 10k to support 20% time of the first-year Assistant Professor’s research time. This additional money will help support a postdoc to help with data QC and analysis.

Personnel: $5,000

Total: $5,000

6. References

1. Manly, J. J. et al. Estimating the Prevalence of Dementia and Mild Cognitive Impairment in the US. JAMA Neurol 79, (2022).

2. Bhole, R. P., Chikhale, R. V. & Rathi, K. M. Current biomarkers and treatment strategies in Alzheimer disease: An overview and future perspectives. IBRO Neuroscience Reports vol. 16 Preprint at https://doi.org/10.1016/j.ibneur.2023.11.003 (2024).

3. Wang, L. et al. Association of COVID-19 with New-Onset Alzheimer’s Disease. Journal of Alzheimer’s Disease 89, (2022).

4. Scott, T. M., Bhadelia, R. A., Qiu, W. Q., Folstein, M. F. & Rosenberg, I. H. Small vessel cerebrovascular pathology identified by magnetic resonance imaging is prevalent in Alzheimer’s disease and mild cognitive impairment: A potential target for intervention. Journal of Alzheimer’s Disease 65, (2018).

5. Klohs, J., Rudin, M., Shimshek, D. R. & Beckmann, N. Imaging of cerebrovascular pathology in animal models of Alzheimer’s disease. Frontiers in Aging Neuroscience vol. 6 Preprint at https://doi.org/10.3389/fnagi.2014.00032 (2014).

6. Bryant, A. G. et al. Cerebrovascular Senescence Is Associated With Tau Pathology in Alzheimer’s Disease. Front Neurol 11, (2020).

7. Laing, K. K. et al. Cerebrovascular disease promotes tau pathology in Alzheimer’s disease. Brain Commun 2, (2020).

8. O’Brien, J. T. & Markus, H. S. Vascular risk factors and Alzheimer’s disease. BMC Med 12, (2014).

9. Sapkota, S. et al. Vascular Risk Predicts Plasma Amyloid β 42/40 Through Cerebral Amyloid Burden in Apolipoprotein e ϵ4 Carriers. Stroke 54, (2023).

10. Conti Filho, C. E. et al. Advances in Alzheimer’s disease’s pharmacological treatment. Frontiers in Pharmacology vol. 14 Preprint at https://doi.org/10.3389/fphar.2023.1101452 (2023).

11. Rius-Pérez, S., Tormos, A. M., Pérez, S. & Taléns-Visconti, R. Vascular pathology: Cause or effect in Alzheimer disease? Neurologia vol. 33 Preprint at https://doi.org/10.1016/j.nrl.2015.07.010 (2018).

12. Le Bihan, D. et al. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology (1988) doi:10.1148/radiology.168.2.3393671.

13. Paschoal, A. M., Leoni, R. F., dos Santos, A. C. & Paiva, F. F. Intravoxel incoherent motion MRI in neurological and cerebrovascular diseases. NeuroImage: Clinical Preprint at https://doi.org/10.1016/j.nicl.2018.08.030 (2018).

14. Le Bihan, D. What can we see with IVIM MRI? Neuroimage 187, 56–67 (2019).

15. Bihan, D. Le & Turner, R. The capillary network: a link between ivim and classical perfusion. Magn Reson Med (1992) doi:10.1002/mrm.1910270116.

16. Zhang, X., Ingo, C., Teeuwisse, W. M., Chen, Z. & van Osch, M. J. P. Comparison of perfusion signal acquired by arterial spin labeling–prepared intravoxel incoherent motion (IVIM) MRI and conventional IVIM MRI to unravel the origin of the IVIM signal. Magn Reson Med (2018) doi:10.1002/mrm.26723.

17. Zhu, G. et al. Comparison of MRI IVIM and MR perfusion imaging in acute ischemic stroke due to large vessel occlusion. International Journal of Stroke (2020) doi:10.1177/1747493019873515.

18. Pasternak, O., Sochen, N., Gur, Y., Intrator, N. & Assaf, Y. Free water elimination and mapping from diffusion MRI. Magn Reson Med (2009) doi:10.1002/mrm.22055.

19. Rydhög, A. S. et al. Separating blood and water: Perfusion and free water elimination from diffusion MRI in the human brain. Neuroimage 156, (2017).

20. Callewaert, B. et al. Intravoxel incoherent motion as a surrogate marker of perfused vascular density in rat brain. NMR Biomed 37, (2024).

21. Iima, M. et al. Characterization of Glioma Microcirculation and Tissue Features Using Intravoxel Incoherent Motion Magnetic Resonance Imaging in a Rat Brain Model. Invest Radiol 49, (2014).

22. Zhang, C. E. et al. Intravoxel Incoherent Motion Imaging in Small Vessel Disease. Stroke (2017) doi:10.1161/strokeaha.116.015084.

23. Lin, H. et al. Associations between microstructural tissue changes, white matter hyperintensity severity, and cognitive impairment: an intravoxel incoherent motion imaging study. Front Aging Neurosci 15, (2023).

24. Tournier, J. D. et al. MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation. NeuroImage Preprint at https://doi.org/10.1016/j.neuroimage.2019.116137 (2019).

25. Jenkinson, M., Beckmann, C. F., Behrens, T. E., Woolrich, M. W. & Smith, S. M. Fsl. Neuroimage 62, 782–790 (2012).

26. Fischl, B. FreeSurfer. Neuroimage 62, 774–81 (2012).

27. Eleftherios, G. et al. Dipy, a library for the analysis of diffusion MRI data. Frontiers in Neuroinformatics Preprint at (2014).

28. Glasser, M. F. et al. The minimal preprocessing pipelines for the Human Connectome Project. Neuroimage 80, 105–24 (2013).

29. Azarpazhooh, M. R. et al. Concomitant vascular and neurodegenerative pathologies double the risk of dementia. Alzheimer’s and Dementia 14, (2018).

Cerebrovascular Disease Underlying White Matter Pathology in Alzheimer’s Disease with and without a History of COVID-19

Cerebrovascular Disease Underlying White Matter Pathology in Alzheimer’s Disease with and without a History of COVID-19

Funding Progress

Funders

Peer Reviews

Related Funding

Request for Proposals - Neuroscience Research Project

Nonprofit Recipient

Topics

DOI