Capillary dysfunction in healthy elderly APOE ε4 carriers with raised brain Aβ deposition

Abstract

INTRODUCTION

Capillary dysfunction, characterized by disturbances in capillary blood flow distribution, might be an overlooked factor in the development of Alzheimer's disease (AD). This study investigated microvascular blood flow in preclinical and prodromal AD individuals.

METHODS

Using dynamic susceptibility contrast magnetic resonance imaging and positron emission tomography, we examined alterations in microvascular circulation and levels of Aβ deposition in two independent cohorts of APOE ε4 carriers.

RESULTS

Capillary dysfunction was elevated in both prodromal and preclinical AD individuals compared to age‐matched controls. Additionally, the prodromal group exhibited higher levels of capillary dysfunction compared to the preclinical group.

DISCUSSION

These findings suggest that capillary dysfunction can be detected at the preclinical stage of AD and indicates a worsening of capillary dysfunction throughout the AD continuum. Understanding the interaction between capillary dysfunction and Aβ could provide insights into the relationship between cardiovascular risk factors and the development of AD.

Highlights

• Alzheimer's disease (AD) is associated with disturbances in microvascular circulation.

• Capillary dysfunction can be detected in preclinical AD.

• As cognitive symptoms progress in prodromal AD, capillary dysfunction worsens.

• Capillary dysfunction may impede the clearance of beta‐amyloid (Aβ).

• Capillary dysfunction might contribute to the development of AD.

1. BACKGROUND

Alzheimer's disease (AD) is a neurodegenerative disease and the most common type of dementia. It is characterized by an abnormal extracellular aggregation of beta‐amyloid (Aβ) protein in the cerebral cortex. ¹ , ² AD‐related pathological changes in the brain can occur up to two decades before any signs of cognitive decline. ³ , ⁴ This makes the early stages of the disease exceedingly difficult to study in life and, consequently, the origin of AD remains elusive. Cortical Aβ aggregation is caused by an imbalance between production and clearance of the protein. However, the question of why this is happening in only some elderly individuals has not been answered. ¹ , ⁵ Alterations in cerebral circulation is another well‐established pathological characteristic of AD. ⁶ , ⁷ , ⁸ , ⁹ Population‐based studies have found that impaired cerebral perfusion is associated with an increased risk of dementia. ¹⁰ Additionally, epidemiological studies have revealed that almost all AD‐related risk factors also affect cerebral perfusion. ¹¹ The association between Aβ aggregation and deficiencies in cerebral perfusion is not fully understood; however, evidence suggests that Aβ oligomers may reduce cerebral blood flow through toxic interaction with contractile capillary pericytes. ¹² , ¹³ Vice versa, a reduction in cerebral blood flow impedes Aβ clearance from brain tissue. Together, this may start a vicious cycle of Aβ build‐up and aggregation and reduced cerebral blood flow. ¹⁴ The initiation of this purported vicious cycle is, however, unknown and requires investigations of the preclinical stages of the disease. The distribution of blood flow through the capillaries is another important aspect of cerebral circulation. Heterogeneous capillary flow, that is, a combination of fast‐ and slow‐flowing capillaries, is normal in a resting condition. ¹⁵ , ¹⁶ Normal functioning capillaries can homogenize the capillary flow in parallel with an increase in blood flow to maintain a high oxygen extraction efficacy during episodes of increased metabolic demand. Minor disruption in the ability to homogenize the capillary flow results in lower oxygen extraction efficacy and, hence, a need for higher blood flow to meet the metabolic demands. ¹⁷ Capillary dysfunction is characterized by progressive decline in the ability to homogenize capillary flow and a progressive increase in capillary blood flow heterogeneity at rest. Severe capillary dysfunction can lead to a state of highly heterogenous capillary flow where blood is functionally shunted through a portion of the capillaries. At this stage, an increase in blood flow does not lead to a parallel increase in oxygen extraction. On the contrary, a higher blood flow increases the functional shunting, thus reducing the time available for oxygen exchange. ¹⁸ , ¹⁹ Evidence of capillary dysfunction has been found in patients with mild cognitive impairment (MCI) before the onset of clinical dementia ²⁰ , ²¹ , ²² ; however, it is unknown what triggers the capillary dysfunction that develops in AD ahead of clinical symptoms.

RESEARCH IN CONTEXT

1. Systematic review: The authors conducted a literature review on microvascular circulation in Alzheimer's disease (AD) using traditional search engines like PubMed. While numerous studies have focused on macroscopic cerebral blood flow in AD, there are limited investigations on the microvascular circulation in vivo. Additionally, no studies were found that specifically examined microvascular circulation in preclinical AD.

2. Interpretation: The findings suggest that cerebral microcirculation is impaired in the early stages of AD and that such disturbances can be detected before the onset of cognitive symptoms.

3. Future directions: Future research should explore longitudinal changes in cerebral microcirculation to understand how these changes progress alongside beta‐amyloid (Aβ) accumulation and cognitive decline. Additionally, there is a need to investigate the mechanistic relationship between disturbances in cerebral microcirculation and Aβ accumulation.

To study preclinical AD in life, identification of individuals at high risk of developing AD is essential. Genetic studies have found that over twenty genes are associated with an increased risk of AD, among which the strongest genetic risk factor is apolipoprotein E (APOE) ε4. ²³ , ²⁴ , ²⁵ Carrying one copy of the APOE ε4 allele increases the risk of AD by 2 to 3 times, while carrying two copies increases the risk by 8 to 12 times. ²³ Untangling the temporal ordering of the pathological insults to the brain in AD is crucial for the success of future treatment strategies. Furthermore, a more thorough understanding of the role of cerebral microvascular changes is needed to elucidate how deficient vascularization interacts with Aβ accumulation and cognitive impairment in the early stages of the disease. The aim of this study was to investigate alterations in cerebral microvascular circulation, including capillary dysfunction, at different stages of early AD. This was done by studying APOE ε4 carriers with MCI as well as healthy elderly APOE ε4 carriers. Both MCI patients and asymptomatic APOE ε4 carriers with evidence of increased cortical Aβ deposition have a higher risk of developing dementia and were considered cases of prodromal and preclinical AD, respectively, in this study.

2. METHODS

2.1. Study subjects

This study comprised two independent cohorts. The first cohort contained patients with prodromal AD and age‐matched healthy controls, denoted the prodromal AD cohort, and the second cohort contained preclinical AD individuals and age‐matched healthy controls, denoted the preclinical AD cohort. Both studies were approved by the local ethics committee and all subjects gave their informed written consent before enrolment in the study.

2.1.1. Prodromal AD cohort

In the prodromal AD cohort, 43 subjects with MCI were recruited from national memory clinics and by advertisement. A history of memory decline for at least 6 months was required for inclusion. Additional inclusion criteria were (1) age between 50 and 85 years; (2) meeting the Petersen criteria for amnestic MCI, ²⁶ that is, memory complaints and objective memory impairment but not demented (no strict memory score cutoff was applied) and able to perform activities of daily living; (3) a Mini‐Mental State Examination (MMSE) score > 24; and (4) no depression, history of stroke, or systemic disease. Additionally, 23 age‐matched healthy control subjects with no overt cognitive impairment or subjective memory complaints were recruited by advertisement with a similar age range and no history of depression, stroke, or systemic disorders. Detailed inclusion criteria are described in Parbo et al. ²⁷ An estimated glomerular filtration rate (eGFR) ≥ 60 was required to safely undergo dynamic susceptibility contrast magnetic resonance imaging (DSC‐MRI). Only APOE ε4 carriers from this cohort were selected for the present study, providing 20 MCI patients and 10 controls. All MCI subjects had an ¹¹C‐Pittsburgh compound B (¹¹C‐PiB)‐positron emission tomography (PET) scan to evaluate their cortical fibrillar Aβ levels. MCI patients with elevated levels of cortical Aβ were considered prodromal AD patients and were included in this study. A flowchart describing the subject inclusion of the prodromal AD cohort is provided in Figure 1A.

FIGURE 1.

Flowchart of subject inclusion. (A) Prodromal AD cohort. (B) Preclinical AD cohort. Aβ, beta‐amyloid; AD, Alzheimer's disease; APOE, apolipoprotein E; MRI, magnetic resonance imaging.

2.1.2. Preclinical AD cohort

Recruitment of healthy elderly subjects by advertisement resulted in finding 61 healthy elderly APOE ε4 carriers. To be included in the study, participants had to (1) be between 50 and 80 years of age; (2) have at least one APOE ε4 allele confirmed by blood genotyping; (3) have no overt cognitive symptoms (MMSE score > 26); and (4) have no significant neurological, psychiatric, or systemic disorders. An eGFR ≥ 60 was required to safely undergo DSC‐MRI. All subjects completed an ¹¹C‐PiB‐PET scan to evaluate their cortical fibrillar Aβ levels. Individuals with elevated levels of cortical Aβ were considered cases of preclinical AD. A flowchart describing the subject inclusion of the preclinical AD cohort is provided in Figure 1B.

2.2. MRI acquisitions

For both cohorts, MRI was performed on the same 3T Skyra scanner (Siemens Healthcare, Erlangen, Germany) using a 32‐channel head coil. All subjects underwent structural T1‐weighted MP2RAGE, ²⁸ T2FLAIR, and perfusion‐weighted DSC‐MRI images. The DSC‐MRI images were acquired with echo‐planar imaging using a spin echo sequence. This spin echo sequence is primarily sensitive to a contrast agent in capillary‐sized vessels and thus reflects microvascular circulation. ²⁹ , ³⁰ Gadobutrol (Gadovist, Bayer HealthCare Pharmaceuticals, Berlin) was used as the contrast agent with a concentration of 0.2 mmol/kg followed by a bolus of 20 mL of saline with an injection rate of 5 mL/s. The molar concentration of the contrast agent was 1 mmol/mL. All T1 and T2FLAIR images were screened for signs of structural abnormalities or brain lesions by a trained radiologist. Due to upgrades in MRI sequences, the scanning protocols for the two cohorts differed slightly as outlined in the following. Due to limitations in image acquisition, whole‐brain coverage was not achievable using DSC‐MRI. Consequently, in the prodromal AD cohort, the superior motor and premotor cortex together with the inferior neocortex and cerebellar regions were outside the MRI field of view (FOV). In the preclinical AD cohort, only a small portion of the superior motor and premotor cortex were outside the MRI FOV.

2.2.1. Prodromal AD cohort

For the prodromal AD cohort, structural T1 was acquired with 1 mm isotropic voxels (TR = 5 s, TE = 2.98 s, TI₁ = 0.7 s, TI₂ = 2.5 s, acquisition matrix 240 × 256 × 176, 4° and 5° flip angles, FOV = 240 × 256 × 176 mm). T2‐FLAIR was acquired with 0.7 × 0.7 × 3.3 mm voxels (TR = 9 s, TE = 117 ms, TI = 2.5 s, 150° flip angle, acquisition matrix 310 × 320 × 45, FOV = 217 × 224 × 148.5 mm). Spin echo DSC‐MRI was acquired with 3 mm isotropic voxels with 1 mm slice‐gap (300 volumes, TR = 1.53 s, TE = 60 ms, 90° flip angle, acquisition matrix 64 × 64 × 19, FOV = 192 × 192 × 76 mm). The voxels were reconstructed to fill the 1 mm slice‐gap, resulting in a voxel size of 3 × 3 × 4 mm.

2.2.2. Preclinical AD cohort

For the preclinical AD cohort, structural T1 was acquired with 0.9 mm isotropic voxels (TR = 6.5 s, TE = 3.46 s, TI₁ = 0.7 s, TI₂ = 2.8 s, acquisition matrix 288 × 288 × 192, 4° and 6° flip angles, FOV = 259.2 × 259.2 × 172.8 mm). T2‐FLAIR was acquired with 0.5 mm × 0.5 mm × 1.0 mm voxels (TR = 5 s, TE = 388 ms, TI = 1.8 s, 120° flip angle, acquisition matrix 512 × 512 × 176, FOV = 256 × 256 × 176 mm). Spin echo DSC‐MRI was acquired with 2.5 mm isotropic voxels (120 volumes, TR = 1.54 s, TE = 60 ms, 90° flip angle, acquisition matrix 86 × 86 × 51, FOV = 215 × 215 × 127.5 mm).

2.3. Perfusion analysis

Pre‐processing and analysis of all DSC‐MRI images were performed using in‐house software ³⁰ using MATLAB (The MathWorks, Inc.) and SPM (The Wellcome Trust Centre for Neuroimaging, University College London). Initially, individual image time series were motion and slice‐time corrected and truncated 60 s after the start of the contrast bolus. Then, automatic selection of arterial input function (AIF) was performed within vascular territories of the left and right middle cerebral artery. Only voxels near the M2 and M3 sections of the middle cerebral artery were considered as possible AIF voxels. Among candidate AIF voxels, the 3 to 7 best AIF voxels (based on AIF characteristics such as time‐to‐peak and area under the curve) were selected and averaged for the final AIF. Parametric maps of relative cerebral blood flow (rCBF), relative cerebral blood volume (rCBV), mean transit time (MTT), and capillary transit time heterogeneity (CTH) were calculated with parametric deconvolution. ³⁰ CTH is defined as the standard deviation of capillary blood transit times and serves as a measure of capillary dysfunction. To be able to compare rCBF and rCBV across individuals, these parameters were normalized to the individual mean value of normal‐appearing white matter (NAWM), as this region is assumed to be relatively spared in early AD. Lastly, voxel‐wise tissue oxygen tension (P_tO₂) was estimated utilizing a biophysical model of brain metabolism. ¹⁷ Given the local microvascular hemodynamics (determined by MTT and CTH measurements), the model estimates the P_tO₂ required to support normal resting cerebral metabolic rate of oxygen, which is assumed to be 2.5 mL/100 g/min. ³¹ Additionally, we assumed a capillary blood volume of 4% and calibrated the biophysical model to yield an oxygen extraction fraction of 0.3 and P_tO₂ of 25 mmHg in NAWM. ³² , ³³

2.4. Amyloid imaging and amyloid status

Individual cortical Aβ load was assessed using ¹¹C‐PiB PET with the same procedure in both cohorts. The scanning was performed on a high‐resolution research tomograph (CTI/Siemens, Knoxville, TN). ¹¹C‐PiB was injected with a mean dose of 391 MBq in the prodromal AD cohort and 389 MBq in the preclinical AD cohort, followed by PET acquisition in list mode 40 to 90 min post injection. After reconstruction, the 60 to 90 min averaged image was used to calculate a standardized uptake value ratio (SUVR) image with cerebellar grey matter as the reference. The resulting SUVR images were then smoothed with a 3 × 3 × 3 mm full width at half maximum (FWHM) Gaussian kernel.

The ¹¹C‐PiB SUVR images were used to identify individuals with elevated cortical Aβ levels. The mean cortical ¹¹C‐PiB uptake was evaluated in a composite region comprising the entire neocortex, excluding the primary visual cortex, the primary somatosensory cortex, the primary motor cortex, and medial temporal regions, as these regions only show late stage Aβ accumulation. ³⁴ A cutoff value of 1.4 SUVR was used to dichotomize subjects. Individuals with a mean ¹¹C‐PiB SUVR within the composite region above the cutoff value were considered to have elevated levels of cortical Aβ deposition (Aβ+).

2.5. Cortical surface mapping

Parametric images from the DSC‐MRI analysis and the ¹¹C‐PiB SUVR images were mapped onto the individual cortical surface prior to the statistical analyses, including evaluation of Aβ levels. Utilizing the high‐resolution T1‐weigthed image, the individual cortical surfaces were segmented using Fast Accurate Cortex Extraction. ³⁵ The middle cortical layer was estimated as the surface between the pial surface (grey matter–cerebrospinal fluid interface) and the white matter surface (white matter–grey matter interface). The parametric images were then sampled onto the corresponding cortical surface. A rigid body coregistration was used to move the cortical surface to parametric native space where the sampling was performed. Individual cortical surfaces were then moved to a common template space using a non‐linear coregistration. ³⁶ Finally, the resulting parametric surfaces were smoothed along the cortex using a 20 mm FWHM geodesic Gaussian kernel. Smoothing of the signal along the cortical surface limits the influence of signal originating from outside the gray matter and avoids smoothing of dissimilar functional regions across sulci. ³⁷

2.6. Cognitive tests

Participants in the preclinical AD cohort underwent cognitive assessment with a number of cognitive tests, including both tests from the standard battery and complex memory tests, to evaluate memory performance. The utilized complex memory tests included verbal memory assessed by Rey Auditory Verbal Learning Test (RAVLT), visual memory assessed by Rey Complex Figure Test (RCFT), and associative memory assessed by Face Name Associative Memory Test (FNAME). In the present study, only delayed recall scores were used from the three memory tests, as these scores have been found to reliably differentiate between Aβ+ and Aβ− individuals. ³⁸ , ³⁹

2.7. Statistical analysis

All statistical analyses were performed using Python 3.7 (Python Software Foundation). Characteristics of participant demographics and the difference between groups in the two cohorts were assessed using two‐tailed unpaired t tests for continuous data and χ ² tests for binary data. Results are presented with a significance level of p < 0.05. The differences in ¹¹C‐PiB SUVR and individual DSC‐MRI perfusion parameters were evaluated using vertex‐wise two‐tailed unpaired t tests adjusted for age. Given the multiple statistical tests, the results were family‐wise error rate (FWER)‐corrected (α = 0.05) using a cluster‐extent‐based thresholding with two primary cluster‐defining thresholds: p < 0.01 and p < 0.001 for ¹¹C‐PiB SUVR, and p < 0.05 and p < 0.01 for the DSC‐MRI parameters. To compare the means of each DSC‐MRI perfusion parameter across the two cohorts, z‐score transformations were performed. For each cohort, the corresponding healthy control group was used as the population mean and standard deviation in the z‐score calculation. The mean of each perfusion parameter was calculated in the cortical area with available perfusion data in all subjects. Consequently, the superior motor and premotor cortex together with the inferior neocortex were excluded due to a limited FOV in the prodromal AD cohort. Direct comparisons of the DSC‐MRI perfusion parameters between the two cohorts were not feasible due to slight changes in the DSC scanning protocols, which resulted in different voxel sizes and hence differences in partial volume effects. Partial volume effects can affect the absolute values of the perfusion parameters, thus a z‐score transformation was necessary to compare the values between cohorts. A two‐tailed unpaired t test was calculated based on the z‐scores between the prodromal AD group and the corresponding healthy control group, between the preclinical AD group and the corresponding healthy control group, and between the prodromal AD group and the preclinical AD group. Finally, a general linear model (GLM) was used to evaluate the correlation between the perfusion parametric maps and the detailed memory test in the preclinical AD cohort. A GLM adjusted for age and mean ¹¹C‐PiB SUVR was fitted at each vertex on the cortical surface and the calculated t‐value was assigned to the corresponding vertex resulting in a t‐value surface map. The correlation results were FWER‐corrected (α = 0.05) using a cluster‐extend‐based thresholding with two levels of primary cluster‐defining threshold: p < 0.05 and p < 0.01.

3. RESULTS

The prodromal AD cohort comprised 13 Aβ+ subjects with MCI, considered cases of prodromal AD, and 8 age‐matched healthy controls though only one had ¹¹C‐PiB PET which was negative. The preclinical AD cohort comprised 20 asymptomatic Aβ+ subjects, considered cases of preclinical AD, and 23 asymptomatic Aβ− subjects, considered healthy controls. All subjects completed DSC‐MRI. Subject demographics of the prodromal AD cohort are presented in Table 1, while subject demographics of the preclinical AD cohort are presented in Table 2. The prodromal AD group had a lower mean cognitive score than the corresponding age‐matched healthy control group (MMSE means: 25.8 and 28.9, p = 0.01). The preclinical AD group was slightly older than the corresponding healthy control group (means: 67.9 and 64.3 years, p = 0.05), but the two groups showed no difference in global cognitive function (MMSE means: 28.8 and 28.5, p = 0.51). However, though asymptomatic, the preclinical AD group performed significantly worse on the comprehensive memory tests: RAVLT delayed recall (means: 7.8 and 10.4, p = 0.01), RCFT delayed recall (means: 13.8 and 18.2, p = 0.02), and FNAME delayed recall (means: 5.6 and 8.5, p = 0.01). In both cohorts, there was a significant difference in the proportion of female/males between the prodromal/preclinical AD group and the corresponding healthy control group.

TABLE 1.

Characteristics of participants of the prodromal AD cohort.

	Prodromal AD (n = 13)	Healthy controls (n = 8)	p value
Age	72.9 [64; 81]	71.4 [62; 78]	0.57 a
Female/Male	3/10	6/2	0.01 b
MMSE	25.8 [22; 30]	28.9 [28; 30]	0.01 a
Antihypertensives (Yes/No)	8/5	1/7	0.03 b
Statins (Yes/No)	6/7	2/6	0.33 b
Mean 11C‐PiB SUVR	2.3 [1.5; 3.0]	n/a	n/a

Notes: Results are presented with range in brackets.

Abbreviations: AD, Alzheimer's disease; MMSE, Mini‐Mental State Examination; n/a, not available; PiB, Pittsburgh compound B; SUVR, standardized uptake value ratio.

^a Unpaired t‐test.

^b χ².

TABLE 2.

Characteristics of participants of the preclinical AD cohort.

	Preclinical AD (n = 20)	Healthy controls (n = 23)	p value
Age	67.9 [56; 75]	64.3 [55; 76]	0.05 a
Female/Male	8/12	19/4	<0.01 b
MMSE	28.8 [27; 30]	28.5 [26; 30]	0.51 a
Antihypertensives (Yes/No)	6/14	3/20	0.17 b
Statins (Yes/No)	5/15	2/21	0.15 b
RAVLT delayed recall	7.8 [0; 15] c	10.4 [0; 14]	0.01 a
RCFT delayed recall	13.8 [4; 26] c	18.2 [8.5; 26] d	0.02 a
FNAME delayed recall	5.6 [0; 12] c	8.5 [0; 12]	0.01 a
Mean 11C‐PiB SUVR	1.7 [1.4; 2.7]	1.3 [1.2;1.4]	<0.01 a

Notes: Results are presented with range in brackets.

Abbreviations: AD, Alzheimer's disease; FNAME, Face Name Associative Memory Test; MMSE, Mini‐Mental State Examination; PiB, Pittsburgh compound B; RAVLT, Rey Auditory Verbal Learning Test; RCFT, Rey Complex Figure Test; SUVR, standardized uptake value ratio.

^a Unpaired t‐test.

^b χ².

^c n = 19.

^d n = 21.

3.1. Amyloid burden on the AD continuum

A comparison between mean cortical ¹¹C‐PiB SUVRs across the three groups with available ¹¹C‐PiB PET (the prodromal AD group, preclinical AD group, and healthy controls from the preclinical AD cohort) is presented in Figure 2 (for detailed information, see Table S1). The mean ¹¹C‐PiB SUVR was significantly higher in the preclinical AD group compared with the corresponding healthy control group in most association cortical regions but not the primary visual cortex, the primary somatosensory cortex, the primary motor cortex, and medial temporal regions. The mean ¹¹C‐PiB SUVR was significantly higher in the prodromal AD group than both the healthy controls and the preclinical AD group, and also targeted association cortical regions, but spared primary cortical areas. ,  

FIGURE 2.

Comparison of mean ¹¹C‐PiB SUVR. Healthy controls and the preclinical AD group from the preclinical AD cohort, and the prodromal AD group from the prodromal AD cohort. Positive t‐values (red colors) indicate significantly higher values in the preclinical AD group compared with healthy controls, and in the prodromal AD group compared with the preclinical AD group, respectively. Statistical maps were family‐wise error rate‐corrected (α = 0.05) using cluster‐extent‐based thresholding with two levels of primary cluster‐defining threshold: p < 0.01 (**) and p < 0.001 (***). Light grey areas indicate regions without any statistically significant difference between the groups. AD, Alzheimer's disease; PiB, Pittsburgh compound B; SUVR, standardized uptake value ratio.

FIGURE 3.

Unpaired t‐test of microvascular perfusion parameters in the prodromal AD cohort. Comparison between 13 prodromal AD individuals and 8 age‐matched healthy controls. Statistical tests were adjusted for age. Positive t‐values (red colors) indicate significantly higher values in the prodromal AD group compared with the healthy control group, and vice versa for negative t‐values (blue colors). Statistical maps were family‐wise error rate‐corrected (α = 0.05) using cluster‐extent‐based thresholding with two levels of primary cluster‐defining threshold: p < 0.05 (*) and p < 0.01 (**). Light grey areas indicate regions without any statistically significant difference between the groups. Dark grey areas indicate regions outside the MRI field of view. AD, Alzheimer's disease; CTH, capillary transit time heterogeneity; MRI, magnetic resonance imaging; MTT, mean transit time; P_tO₂, tissue oxygen tension; rCBV, relative cerebral blood volume; rCBF, relative cerebral blood flow.

FIGURE 4.

Unpaired t‐test of microvascular perfusion parameters in the preclinical AD cohort. Comparison between 20 preclinical AD individuals and 23 age‐matched healthy controls. Statistical tests were adjusted for age. Positive t‐values (red colors) indicate significantly higher values in the preclinical AD group compared with the healthy control group, and vice versa for negative t‐values (blue colors). Statistical maps were family‐wise error rate‐corrected (α = 0.05) using cluster‐extent‐based thresholding with two levels of primary cluster‐defining threshold: p < 0.05 (*) and p < 0.01 (**). Light grey areas indicate regions without any statistically significant difference between the groups. Dark grey areas indicate regions outside the MRI field of view. AD, Alzheimer's disease; CTH, capillary transit time heterogeneity; MRI, magnetic resonance imaging; MTT, mean transit time; P_tO₂, tissue oxygen tension; rCBV, relative cerebral blood volume; rCBF, relative cerebral blood flow.

3.2. Microvascular dysfunction in individuals with AD pathology

The prodromal AD group showed significant disturbances in microvascular circulation compared with the corresponding healthy control group (Figure 3 and Table S2). Specifically, disturbances in perfusion parameters linked to the distribution of blood flow, that is, elevated MTT and CTH together with reduced P_tO₂, were found in widespread cortical areas (p < 0.01). Only smaller areas, primarily located in the cingulate cortex and precuneus, showed reduced rCBV and rCBF in the prodromal AD group. A similar pattern of altered microvascular circulation was detected in the preclinical AD group (Figure 4 and Table S3). Significantly elevated levels of MTT and CTH, together with reduced P_tO₂, were found in the preclinical AD group in extensive areas of the cortex compared with the corresponding healthy control group (p < 0.05), although these were less significant than in the prodromal AD cohort. Likewise, the preclinical AD group showed smaller areas of reduced rCBF and rCBV in occipital and temporal regions.

3.3. Microvascular dysfunction on the AD continuum

Figure 5 displays a comparison between the mean z‐score of each perfusion parameter across the four groups. No significant difference was found in the mean z‐scores of rCBF and rCBV between the prodromal AD group and the corresponding healthy control group or between the preclinical AD group and the corresponding healthy control group. Furthermore, no difference was found between the prodromal AD and preclinical AD groups. In contrast, a significant difference was found in mean z‐scores of CTH, MTT, and P_tO₂ between the prodromal AD and corresponding healthy control group (p < 0.01 for MTT and P_tO₂, p < 0.05 for CTH) and between the preclinical AD and corresponding healthy control group (p < 0.05). Additionally, a significant difference was found in CTH, MTT, and P_tO₂ between the prodromal AD and preclinical AD groups (p < 0.01).

FIGURE 5.

Unpaired t‐test of mean z‐score of microvascular perfusion parameters across cohorts. Comparison between the preclinical AD group (n = 20) and the corresponding control group (n = 23) from the preclinical AD cohort, and the prodromal AD group (n = 13) and the corresponding control group (n = 8) from the prodromal AD cohort. The mean z‐score was calculated in the area with perfusion data available in all subjects, as indicated by the cortical surface in the lower right corner. For each cohort, the corresponding healthy control group was used as the population mean and standard deviation to calculate individual z‐scores. *p < 0.05, **p < 0.01, ns = p > 0.05. AD, Alzheimer's disease; CTH, capillary transit time heterogeneity; MTT, mean transit time; Prec. AD, preclinical AD; Prod. AD; prodromal AD; HC, healthy controls; P_tO₂, tissue oxygen tension; rCBV, relative cerebral blood volume; rCBF, relative cerebral blood flow.

3.4. Reduced microvascular blood flow correlates with memory impairment

In the preclinical AD cohort, a widespread positive correlation was found between microvascular rCBF and the memory scores of RAVLT delayed recall and FNAME delayed recall after adjusting for age and mean ¹¹C‐PiB SUVR (Figure 6 and Table S4). This correlation was most significant in superior parietal and temporal regions (p < 0.01). Only a small cluster (p < 0.05) in the posterior temporal lobe of the left hemisphere showed correlation between microvascular rCBF and RCFT delayed recall. No apparent correlation was found between the memory scores and any other perfusion parameters (rCBV, MTT, CTH, or P_tO₂) after adjusting for age and mean ¹¹C‐PiB SUVR.

FIGURE 6.

Correlation between cerebral blood flow and memory. Cognitive scores of RAVLT delayed recall, FNAME delayed recall, and RCFT delayed recall were used from the preclinical AD cohort (n = 42). The correlations were adjusted for age and ¹¹C‐PiB SUVR. Positive t‐values (red colors) indicate a significant positive correlation. Statistical maps were family‐wise error rate (FWER)‐corrected (α = 0.05) using cluster‐extent‐based thresholding with two levels of primary cluster‐defining threshold: p < 0.05 (*) and p < 0.01 (**). The correlation between cerebral blood flow and each cognitive score is shown using the average value within the most significant FWER‐corrected cluster with a primary cluster‐defining threshold of p < 0.01 (indicated by the black arrows). Light grey areas indicate regions without any statistically significant correlation. Aβ, beta‐amyloid; AD, Alzheimer's disease; FNAME, Face Name Associative Memory Test; PiB, Pittsburgh compound B; rCBF, relative cerebral blood flow; RAVLT, Rey Auditory Verbal Learning Test; RCFT, Rey Complex Figure Test; SUVR, standardized uptake value ratio.

4. DISCUSSION

In the present study, we evaluated the microvascular circulation in two independent APOE ε4 cohorts with groups at different stages on the AD continuum. The first cohort contained patients with MCI and elevated cortical Aβ, considered cases of prodromal AD, while the second cohort contained asymptomatic individuals with elevated cortical Aβ, considered to be cases of preclinical AD. We found disturbances in perfusion parameters linked to the distribution of microvascular blood flow (capillary dysfunction) in both the prodromal and preclinical AD groups compared with their corresponding healthy control groups. A comparison between the prodromal AD group and the preclinical AD group showed a higher degree of microvascular flow disturbances in the prodromal AD cases. A positive correlation was found between rCBF and memory scores after adjusting for age and mean ¹¹C‐PiB SUVR. The other perfusion parameters, rCBV, MTT, CTH, and P_tO₂, did not correlate with the memory scores.

The major finding of this study is that capillary dysfunction can already be detected at the preclinical stage of AD before the onset of overt cognitive symptoms. As cognitive symptoms arise at the prodromal stage, capillary dysfunction worsens. This tendency is highlighted by the findings that the preclinical AD group showed higher levels of capillary dysfunction than the healthy control group but lower than the prodromal AD group. This trajectory is further emphasized by previously reported results of progressive worsening of capillary dysfunction in prodromal AD patients over a 2‐year period (mixed APOE genotype). ²⁰ As indicated by the extended flow‐diffusion model, ¹⁷ the ability of the microvasculature to decrease blood flow heterogeneity is crucial to maintaining high oxygen extraction efficacy during episodes of increased metabolic demand. Inadequate blood flow homogeneity may cause local hypoxia in parts of the capillary network due to functional shunting of blood through fast‐flowing capillaries. ¹⁸ Chronic hypoxia can in turn cause increased production of Aβ by modulating the metabolism of the amyloid precursor protein. ¹² , ⁴⁰ Additionally, hypoxia‐inducible factor 1‐α can mediate an up‐regulation of beta‐site amyloid precursor protein cleaving enzyme 1, which further increases Aβ production. ⁴¹ , ⁴² Oligomers of Aβ may reinforce capillary dysfunction as they are toxic to contractile capillary pericytes. ¹² , ¹³ Conversely, capillary dysfunction may reduce the clearance of Aβ from the tissue due to lower blood flow in contracted capillaries. Altogether, chronic hypoxia and the effects of Aβ aggregates on capillaries may provide a link between capillary dysfunction and the development of AD. This is supported by previous findings of capillary dysfunction correlating with a higher cortical Aβ load in prodromal AD patients. ⁴³

Several pathways have been suggested as being involved in the clearance of Aβ from the brain, ⁴⁴ and some of these clearance pathways involve capillary endothelial cells and pericytes. ⁴⁵ , ⁴⁶ Damage to these cells is also thought to be involved in capillary dysfunction; however, the exact mechanistic relationship between capillary dysfunction, Aβ clearance, and dysfunctional capillary endothelial cells and pericytes remains to be studied. In addition, it is unclear whether the observed disturbances in the microvascular hemodynamic is sufficient to cause tissue hypoxia in preclinical or prodromal AD. This also warrants further investigation.

Indications of capillary dysfunction in other pathological conditions, such as post stroke, ⁴⁷ traumatic head injury ⁴⁸ and Parkinson's disease, ⁴⁹ indicate that capillary dysfunction is not a phenomenon specific to AD. Instead, it may serve as an underlying pathological mechanism reinforcing other ongoing pathological processes. Multiple hypotheses of AD pathogenesis have been proposed; however, the exact mechanisms remain elusive. Likewise, the role of capillary dysfunction in the early development of AD is not well established. Increasing evidence now suggests that AD may be caused by a range of accumulative risk factors, including cardiovascular, genetic, and environmental factors, rather than one specific disease mechanism. ⁵⁰ , ⁵¹ Measurements of capillary dysfunction could provide a surrogate marker of cardiovascular risk‐factors; however, other AD‐specific risk factors (genetic or environmental) might be necessary to start a pathological circle of progressive capillary dysfunction, chronic hypoxia, and Aβ accumulation leading to downstream tau tangles and neuronal dysfunction. Our findings tie in well with observations of altered cerebral perfusion in young APOE ε4 carriers ¹⁸ , ⁵² , ⁵³ and suggest that capillary dysfunction may develop over many years while promoting disease‐specific pathological processes such as Aβ accumulation in AD.

No correlation was detected between capillary dysfunction and memory after adjusting for effects of age and Aβ load. This further indicates an indirect role of capillary dysfunction on cognition in AD. By contrast, reduced rCBF correlated with lower performance on the comprehensive memory test performed in the preclinical AD cohort. Multiple studies have found reductions in CBF correlating with worse cognitive performance in both diseased individuals and in the general population. ¹⁰ , ⁵⁴ , ⁵⁵ , ⁵⁶ Besides being toxic to capillary pericytes, evidence also suggest that Aβ induces arteriolar dysregulation by acting as a potent vasoconstrictor. ⁵⁷ This, in turn, causes loss of dynamic CBF regulation which could contribute to cognitive impairment. ⁵⁸ Furthermore, stalling of capillary blood flow by neutrophils has been shown to be associated with reduced CBF in an AD mouse model. ⁵⁹ Blocked capillary segments, in turn, would also contribute to increased capillary dysfunction.

Studies show robust evidence of reduced CBF in AD as well as in at‐risk individuals. ⁹ , ¹⁰ However, in the present study we only detected a small degree of reduction in rCBF in our prodromal and preclinical AD cases. To be able to compare the parametric rCBF and rCBV acquired with DSC‐MRI across individuals, a normalization was necessary. We used NAWM as a reference, that is, white matter without hyperintensities on the T2FLAIR image. This choice was based on the assumption of relatively fewer pathological changes in white matter compared to grey matter in AD. However, possible parallel reductions in white matter rCBF and rCBV would have masked the effects of grey matter rCBF and rCBV reduction after normalization. ⁶⁰ Together with a relatively small sample size, this might explain the divergent results.

We were unable to evaluate the Aβ level of most of the healthy control subjects in the prodromal AD cohort, as only one completed an ¹¹C‐PiB PET scan. Assuming one third of the control group in the prodromal AD cohort were Aβ+, we might expect elevated levels of CTH and MTT together with decreased P_tO₂ in these healthy control subjects affecting the overall group mean. However, despite this possible confound, we observed widespread significant difference between the prodromal AD group and the healthy control group. This indicates a progressive worsening of capillary dysfunction throughout the AD continuum; however, further longitudinal studies are needed for clarification.

Preclinical AD is defined as individuals with elevated amyloid burden but without any overt cognitive impairment. In order to reduce the number of PET scans needed to recruit a sufficient number of preclinical AD cases, only APOE ε4 carriers were recruited for screening with ¹¹C‐PiB PET as APOE ε4 carriers have a 2 to 3 times increased risk of AD. In addition to increasing the risk of AD, the APOE ε4 genotype has also been shown to affect cerebral perfusion. Multiple possible explanations for the connections between APOE ε4 and cerebrovascular deficiencies have been presented. First, evidence suggests a direct interaction of APOE ε4 on Aβ accumulation, ²⁵ , ⁶¹ , ⁶² which, as previously outlined, can damage capillary pericytes ¹³ , ⁶³ and cause arteriolar dysregulation. ⁵⁷ Second, APOE ε4 can directly affect cerebral perfusion by disrupting lipid homeostasis by raising total cholesterol, leading to a higher amount of atherosclerosis and subsequent damage to capillaries. ²⁵ , ⁶⁴ Finally, APOE ε4 might indirectly affect cerebral perfusion through promoting microglial activation and other inflammatory processes. ⁶⁵ , ⁶⁶ Future studies will include APOE ε3 carriers to evaluate whether the findings of the present study are representative of preclinical AD regardless of APOE genotype.

The present study has a number of limitations. First, the number of subjects in each group was relatively low, especially in the prodromal AD cohort, thus limiting the statistical power. However, the similar pattern of capillary dysfunction in the two independent cohorts strengthens the findings of the study. Additionally, the perfusion‐weighted DSC‐MRI protocol differed between the two cohorts, resulting in a relatively large difference in voxel sizes. Due to the complexity of susceptibility contrast as it pertains to the AIF used in our analyses, this difference may introduce systematic differences in AIF and consequently in perfusion parameters. Therefore, direct comparison of the perfusion parameters between the two cohorts was not reported. The DSC‐MRI images of the prodromal AD cohort did not cover the entire brain, and the relatively large voxel sizes of the imaging sequences make the parametric images susceptible to partial volume effects. Finally, no detailed information on cardiovascular risk factors or medical history was available for either cohort, which could have potentially influenced the development of capillary dysfunction.

In conclusion, this study has presented evidence of capillary dysfunction in prodromal AD patients compared with healthy controls which was replicated in preclinical AD individuals compared with healthy controls in a separate cohort. Comparing the level of capillary dysfunction across the two cohorts showed a tendency of worse capillary dysfunction in prodromal AD patients compared with preclinical AD individuals. No correlation was found between capillary dysfunction and memory performance, suggesting an indirect role of capillary dysfunction on clinical outcomes.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information

CONSENT STATMENT

Both studies used in this article were approved by the local ethics committee and all subjects gave their informed written consent before enrolment in the study.

Supporting information

Supporting Information

Supporting Information

Madsen LS, Kjeldsen PL, Ismail R, et al. Capillary dysfunction in healthy elderly APOE ε4 carriers with raised brain Aβ deposition. Alzheimer's Dement. 2024;20:459–471. 10.1002/alz.13461