Cerebral hypoperfusion, brain structural integrity, and cognitive impairment in older APOE4 carriers

Abstract

Cerebral blood flow (CBF) deficits, cognitive decline, and brain structural changes have been reported in older adults with and without apolipoprotein E-e4 (APOE4)-related risk for dementia. However, it remains unclear whether brain structural changes mediate the effects of hypoperfusion on cognitive impairment in APOE4 carriers and non-carriers. We studied 166 (60–89 years) APOE4 carriers (ε3/ε4 or ε4/ε4) and APOE3 homozygotes (e3/e3) with and without cognitive impairment by clinical dementia rating (CDR) and neuropsychological testing. Pseudocontinuous arterial spin-labeling-MRI assessed regional CBF, and T1-anatomical and diffusion-MRI assessed structural integrity. Mediation analyses examined relationships among grey matter CBF, grey matter volume, and white matter integrity in regions underlying impairment in distinct cognitive ability domains. APOE4 carriers with global/memory impairment (CDR 0.5) exhibited decreased CBF in the posterior cingulate, decreased grey matter volume in the hippocampus, parahippocampal gyrus, and posterior cingulate, and decreased white matter integrity in the cingulum relative to APOE4 carriers with no impairment (CDR 0). Mediation analysis in APOE4 carriers indicated decreased posterior cingulate CBF effects on global/memory impairment were mediated by decreased cingulum integrity. In the combined APOE4 and APOE3 carriers sample, there were direct effects of frontal and inferior parietal CBF and superior longitudinal fasciculus integrity on attention/executive impairment. There were also direct effects of left inferior frontal CBF on language impairment. Findings suggest links between hypoperfusion and brain structural integrity underlying global/memory impairment in APOE4 carriers. Independent CBF relationships with structural integrity are also identified across genotypes and impairment domains.

Introduction

Age-related cerebrovascular changes are associated with grey matter atrophy, white matter injury, cognitive decline, and dementia in older adults [1, 2]. In particular, deficits in cerebral blood flow (CBF) occur with aging [3, 4] and may lead to tissue hypoperfusion and a mismatch in neuronal metabolic supply versus demand, possibly contributing to brain injury and neurodegeneration [5]. Consistent with this hypothesis, decreased CBF on perfusion MRI is among the earliest and most predictive biological changes associated with cognitive decline and dementia in older adults [6–8]. The role of cerebral hypoperfusion injury in carriers of the apolipoprotein E-ε4 (APOE4) risk gene for Alzheimer’s disease (AD) and dementia is less established.

The APOE4 allele is the single greatest genetic risk factor for Alzheimer’s disease (AD) and dementia [9, 10], and also conveys increased risk for brain vascular disease [11]. Experiments in humanized APOE4 transgenic animals suggest that vascular effects of APOE4 include cerebral hypoperfusion injury [12, 13]. However, research to date in humans suggests a complex pattern of age-related changes in CBF in APOE4 carriers. For example, some studies in middle-aged and older cognitively unimpaired APOE4 carriers find increased regional CBF relative to non-carriers [3, 14–16]. It remains unclear whether this early hyperperfusion represents a pathological or compensatory vasodilation, and findings are mixed regarding the association between early CBF differences and better, worse, or equivalent cognition in APOE4 carriers [3, 15–21]. In contrast, others have reported decreased CBF in older APOE4 carriers relative to non-carriers [22], and that decreased CBF with aging may be more strongly associated with worsening cognition in APOE4 carriers relative to non-carriers, particularly in those with cognitive impairment [17, 22–24]. However, not all studies have shown a consistent pattern of perfusion changes in APOE4 carriers with cognitive impairment [16, 21, 25], and most research has not focused on tying regional hypoperfusion patterns to structural changes underlying specific cognitive ability domains.

One study focusing specifically on memory regions found decreased CBF in older APOE4 carriers with memory impairment [24], a finding that was particularly pronounced when evidence of structural volume loss was also present in those memory regions [23]. These findings could suggest that age-related deficits in CBF may impact cognition through changes in brain structural integrity in APOE4 carriers. We hypothesize that age-related decline in CBF in APOE4 carriers may cause cognitive impairment through hypoperfusion injury to brain grey and white matter structures underlying important cognitive ability domains. To address this question, the present study investigates regional CBF differences between older APOE4 carriers and APOE3 homozygotes with and without cognitive domain impairment, and whether differences in related grey and white matter structural integrity mediate the effects of hypoperfusion on cognitive impairment in relevant ability domains.

Methods

Participants

All 166 participants were recruited from the community and the USC Alzheimer’s Disease Research Center (ADRC) if they were living independently and between the ages of 55 and 89. Study exclusions were a history of clinical stroke, dementia, moderate-to-severe traumatic brain injury, major psychiatric or neurologic illness, substance abuse, organ failure or major systemic illness, or active medications or other medical conditions that could impact cognitive function, as well as contraindications for contrast-enhanced brain MRI. Participants were not excluded based on the presence of stable, controlled cardiovascular risk factors, such as hypertension, hyperlipidemia, coronary artery disease, or type 2 diabetes. The study and procedures were approved by the USC ADRC Institutional Review Board indicating compliance with all ethical regulations. Informed consent was obtained from all participants before study enrolment.

Brain MRI

All MRI scans were acquired using 3 T scanner at Mark and Mary Stevens Neuroimaging and Informatics Institute of USC (Siemens Prisma 32-channel head receive coil and body transmit coil scanner). All 166 participants underwent a high-resolution 3D T1-weighted magnetization-prepared rapid acquisition gradient echo (MPRAGE) MRI T1-weighted for structural imaging with the following parameters: TR/TE = 7.37/3.05 ms; TI = 400 ms; NEX = 1; flip angle = 11°; FOV = 256 × 256 mm with isotropic 1-mm voxel size; and 196 slices. T1-weighted image denoising, bias correction, and brain mask extraction were performed using the Rican correction method [26], N4BiasFieldCorrection command [27], and antsBrainExtraction.sh command [28] respectively. The T1 image was registered to the b = 0 s/mm² and the fractional anisotropy (FA) images [29] using nonlinear Syn ANTs multivariate option with two target images (b0 and FA) and the moving T1 image. The white matter, grey matter, and cerebrospinal fluid partial volume maps were extracted with fast from the FSL package [30].

Perfusion MRI

All 166 participants underwent cerebral perfusion MRI with a 3D gradient and spin-echo pseudo-continuous arterial spin labeling (pCASL) sequence with background suppression. Parameters were as follows: 2.5 mm³ isotropic resolution, matrix 96 × 96, 48 slices, 4 segments, TR/TE = 4300 ms/36.8 ms, 120° flip angle, label duration = 1500 ms, post-labeling delay = 2000 ms, labeling efficiency = 0.73. The pCASL scans were pre-processed using the ASLtbx pipeline, implemented in SPM12 within MATLAB [31, 32]. Pre-processing steps for pCASL scans included [33] motion correction, co-registration to individual subject’s structural T1-weighted image, spatial smoothing with a 6-mm full-width at half-maximum Gaussian kernel, and tag-control subtraction resulting in 15 tag-control pairs for each subject with values for absolute CBF (mL/100 g tissue/min). All CBF images were thresholded below 10 or above 150 mL/100 g/min to exclude CBF outside the expected physiological range of grey matter [33, 34]. Tag-control pairs were warped to MNI space and averaged to create mean CBF maps for each subject. Resulting mean CBF maps were visually inspected for quality and gross abnormalities (i.e., large signal dropout). Partial volume correction was performed by applying subject-specific grey matter masks derived from the grey matter tissue class segmentation of T1-weighted structural images [35]. Segmented grey matter maps were thresholded at 0.3, binarized, and multiplied by the mean CBF maps to ensure CBF was limited to grey matter. Regional CBF values were extracted from all ROIs using the AAL3 atlas and were normalized by residualizing to the precentral gyrus using linear regression analysis. All CBF data are reported as standardized residual z scores.

Diffusion MRI

A subset of 99 participants from the overall sample of 166 undergoing perfusion MRI also underwent high-angular diffusion tensor imaging (DTI) to acquire diffusion weighted images (DWI) with a spin-echo EPI sequence with the following parameters: TR/TE = 3230/89.2 ms, flip angle = 78°, matrix = 140 × 140, 92 slices of 1.5 mm with 1.25 × 1.25-mm voxel size, multiband factor = 4, echo spacing = 0.78 ms, BW = 1700 Hz/Px, partial Fourier factor 0.75. After post-processing and quality control, 96 out of the 99 participants undergoing DTI-MRI had diffusion images of sufficient quality for analysis. First, DWI were denoised based on the dwidenoise from MRtrix3 using the MP-PCA method [36]. For Eddy/Topup, a bet prelim DWI task is first extracted using FSL’s bet method [37] and Eddy/Topup was run only within the mask to correct the brain deformation induced by the magnetic field susceptibility artifacts. Next, eddy was used to correct eddy-currents, motion artifacts and to perform slice-wise outlier detection and correction. Phase-encoding direction was AP, and a reverse phase encoded b = 0 s/mm² image was provided by using an additional PA image. The topup command was run on the b = 0 s/mm² and reversed phase encoded b = 0 s/mm² images to extract the deformation field. The deformation field was applied, and the eddy command was performed using the topup output. A brain extraction mask was extracted from the b = 0 s/mm² image (using bet) and applied to the whole brain DWI. An N4 bias field correction was performed on the b = 0 s/mm² image using the ANTs command N4BiasFieldCorrection. The DWI was cropped to reduce the bounding box and resampled to 1 mm to match the T1 resolution. Diffusion reconstruction occurred using b values of 1500 s/mm². The metrics computed from the diffusion tensor reconstruction [38, 39].

White matter tractography analysis was conducted with DWI images processed using the TractoFlow pipeline [40] that uses Nextflow [41] and singularity [42]. The pipeline computes DTI maps, and a whole brain tractogram and calls different functions from various neuroimaging software packages, namely FSL [43], MRtrix3 [44], ANTs 28, and DIPY [45]. White matter bundles were extracted using the RecoBundles framework as implemented in DIPY [45]. RecoBundles transforms the tractogram to MNI space using streamline-based linear registration then extracts the tracts from the registered tractogram using the MNI atlas “Advanced_Atlas_of_80_Bundles_in_MNI_space” (https://figshare.com/articles/dataset/Advanced_Atlas_of_80_Bundles_in_MNI_space/7375883) and extracts the uncinate fasciculus (UF) (left and right), the cingulum (left and right), and the superior longitudinal fasciculus (SLF) tracts. Extracted FA values of segments along the tracts were used in statistical analysis. After quality control analysis of processed DTI data, the pipeline failed to provide any usable data on three participants, yielding a subset of n = 96 with usable DTI data for cingulum and SLF tracts, and a subset of n = 94 with usable UF tract data.

APOE genotyping

Briefly, DNA was extracted from buffy coat followed by APOE genotyping by polymerase chain reaction (PCR)-restriction fragment length approach [46]. Participants were stratified based on APOE genotype as APOE4 carriers (ε3/ε4 and ε4/ε4) and homozygote APOE3 allele (ε3/ε3).

Cognitive and neuropsychological evaluation

Each participant underwent clinical dementia rating (CDR) evaluation and standardized clinical interview as part of health history assessment, a physical examination, and neuropsychological testing according to unified data set (UDS) (version 2.0/3.0) procedures. For global/memory domain impairment, impairment was defined as CDR 0.5 and unimpaired as CDR 0. For attention/executive and language domain impairments, we applied previously validated neuropsychological criteria for cognitive domain impairment versus cognitively unimpaired [47]. For attention/executive domain impairment, impairment was defined as having two or more attention/executive domain test scores falling > 1 standard deviation below demographically-corrected normative values. Attention/executive test scores included Trails A, Trails B, and digit/number span. For language domain impairment, impairment was defined as having two or more language domain test scores falling > 1 standard deviation below demographically corrected normative values. Language domain test scores included animals, F-A-S, and confrontation naming (Multilingual Naming Test/Boston Naming Test). All normative values were corrected for age, sex, and education using publicly available UDS version2/3-derived regression equations.

Vascular risk factors (VRFs)

Each participant underwent a physical examination, blood tests, and interviews to determine presence of VRFs. The presence of VRFs were defined based on classification from the Framingham Stroke Risk Profile and included a history of cardiovascular disease (heart failure, angina, stent placement, coronary artery bypass graft, intermittent claudication), hypertension, hyperlipidemia, type 2 diabetes, atrial fibrillation, and transient ischemic attack or minor stroke. The total burden was defined by the sum of these specific risk factors. Based on prior studies linking 2 + VRFs vs. 0–1 VRFs to cerebrovascular pathology [48, 49], participants were grouped by total VRF burden of 0–1 vs. 2 +.

Statistical analysis

All study variables were screened for outliers through visual inspection of variable distributions and removal of influential outliers greater than or equal to +/− 3 standard deviations from the overall sample mean. This led to removal of 1 hippocampal CBF value, 1 parahippocampal gyrus value, 1 uncinate fasciculus FA value, 1 cingulum FA value, and 1 superior longitudinal fasciculus value. Based on prior studies [23, 24], our a priori hypothesis was that cognitively unimpaired APOE4 carriers would exhibit decreased CBF relative to APOE3 homozygotes, and cognitively impaired APOE4 carriers would exhibit decreased CBF relative to cognitively unimpaired APOE4 carriers, and that APOE4 carriers with cognitive impairment would show decreased CBF relative to cognitively impaired APOE3 homozygotes. This hypothesis drove our a priori planned comparisons of APOE4 carriers versus APOE3 homozygotes with and without cognitive impairment in regions involved in memory and global cognitive function, including the hippocampal, parahippocampal and posterior cingulate regions. Pairwise group comparisons from a 2 × 2 analysis of covariance (ANCOVA), comparing APOE status (APOE4 vs. APOE3/3) × cognitive status (CDR 0.5 vs. 0), controlling for age, sex, education, and VRF burden (2 + vs. 0–1 VRFs) were utilized as previously validated [48, 49]. Findings were also confirmed by hierarchical logistic regression analysis of regional CBF as a predictor of cognitive status (CDR 0.5 vs. 0) with same covariates.

To determine whether cerebrovascular changes may be related to changes in grey and white matter structural integrity underpinning memory and global cognitive function, the same approach was applied to evaluation of group differences (2 × 2 ANCOVA: APOE status × cognitive status) in hippocampal and parahippocampal volume, as well as white matter integrity (DTI FA values) in the connected uncinate fasciculus. Posterior cingulate gyrus volume and white matter integrity in the connected cingulum tract were also evaluated. All analyses controlled for age, sex, education, and VRF burden. For brain regions showing differences in CBF or structural integrity related to genotype and/or cognitive impairment, mediation analysis was used to confirm direct effects of regional CBF independent of brain structural integrity and to test whether indirect effects of regional CBF were mediated by brain structural integrity on cognitive status (CDR 0.5 vs. 0), controlling for age, sex, education, and VRF burden. Exploratory analyses using the same approach examined CBF, and grey and white matter integrity in connected regions and tracts underpinning attention/executive and language domains, and both main effects and pairwise comparisons were explored. Attention/executive and language domain impairments were age-, sex-, and education-corrected, and all models additionally corrected for VRF burden.

To investigate regional FA differences along the white matter tracts, we used the BUndle ANalytics (BUAN) [50] framework alongside custom scripts. Briefly, the FA values for all segments of the tracts were extracted as an.h5 object. These values were then used as predicted values in a linear mixed effects model with group, age, sex, education, and vascular risk factors as predictors. Group was a variable representing group membership during the comparison between the respective groups (for example APOE4 vs. APOE3). A p value per segment was obtained (uncorrected) for each tract. p values < 0.05 were plotted as red segments over the respective tract. Mean FA values over the entire tracts were also extracted and used in the swarmplot figures.

All analyses were 2-tailed with significance set at p < 0.05. False discovery rate (FDR)-correction (Benjamini-Hochberg) was applied for all a priori planned comparisons to address multiple comparisons. Follow-up confirmatory, exploratory, and BUAN tract segment analyses were not multiple comparisons corrected. Mediation analysis using the Hayes process macro in R findings are presented as the regression coefficient (unstandardized β) with Bootstrap 95% confidence interval (CI). Statistical analyses were performed in R and SPSSv29.

Results

Demographic and clinical characteristics of the sample are displayed in Table 1. For regions underlying global/memory function, no CBF differences by genotype or CDR status were observed in the hippocampus or parahippocampal gyrus. In the posterior cingulate gyrus, global/memory impaired APOE4 carriers exhibited significantly decreased CBF relative to unimpaired APOE4 carriers, p = 0.002. Additionally, unimpaired APOE4 carriers showed a small but statistically significant increase in CBF in the posterior cingulate gyrus relative to unimpaired APOE3 homozygotes (p = 0.02); however, this finding did not survive FDR correction for multiple comparisons (Fig. 1a–d). Results in APOE4 carriers were confirmed by logistic regression analysis showing only posterior cingulate CBF predicts global/memory impairment by CDR, p = 0.007 (Fig. 1e). All models controlled for age, sex, education, and VRF burden.

Table 1.

Demographic and clinical characteristics of the sample

APOE genotype		APOE3/3	APOE3/4	APOE3/3	APOE3/4
CDR score		0	0	0.5	0.5
pCASL-MRI N = 166	n	70	53	19	24
MPRAGE-MRI N = 166	n	70	53	19	24
DTI-MRI (Cingulum) Subset of n = 95	n	42	34	8	11
DTI-MRI (UF) Subset of n = 93	n	41	33	8	11
DTI-MRI (SLF) Subset of n = 95	n	42	34	8	11
Age, years	Mean (SD)	69.8 (6.8)	68.2 (7.3)	72.0 (8.1)	71.5 (6.7)
Female	n (%)	29 (41.4)	21 (39.6)	10 (52.6)	11 (45.8)
Education, years	Mean (SD)	16.6 (2.4)	16.3 (2.1)	15.4 (3.3)	15.9 (2.5)
Attention/executive impairment	n 0/1	49/5	38/4	9/2	6/5
Language impairment	n 0/1	50/5	38/5	10/2	7/9
Vascular risk factor burden	n 0–1/2 +	41/29	27/26	9/10	11/13

Of the N = 166 with pCASL-MRI data, a subset of n = 99 also underwent diffusion imaging and n = 96 of these participants had usable cingulum and SLF data, and n = 93 had usable UF data. One outlier was then removed from each tract to yield final numbers displayed. Numbers for the participant subset in the final analysis are presented above after outlier removal (see Methods for details). CDR, clinical dementia rating; pCASL, pseudocontinuous arterial spin-labeling; DTI, diffusion tensor imaging; MRI, magnetic resonance imaging; SD, standard deviation; UF, uncinate fasciculus; SLF, superior longitudinal fasciculus

Fig. 1.

a Representative CBF images from ROIs related to global/memory impairment are displayed, including hippocampus (HC), parahippocampal gyrus (PHG), and posterior cingulate gyrus (PCG). Scale bar is − 1 to + 2 residualized z score. b–d Regional CBF values (residualized z scores) are displayed in violin plots with median and interquartile range and compared by APOE4 status (APOE3 homozygotes [blue] vs. APOE4 carriers [red]) and global/memory impairment by clinical dementia rating (CDR) global score (0 vs. 0.5): CDR 0 APOE3 (n = 70), CDR 0 APOE4 (n = 53), CDR 0.5 APOE3 (n = 19), CDR 0.5 APOE4 (n = 24). e Findings from confirmatory hierarchical logistic regression analysis are shown with logistic regression coefficients displayed on the x axis expressed in log-odds units, and error bars representing standard errors. All analyses controlled for age, sex, education, and VRF burden

For structural grey matter measures in global/memory regions, we observed significantly decreased hippocampal (p = 0.00003), parahippocampal (p = 0.002), and posterior cingulate (p = 0.02) volumes in global/memory impaired APOE4 carriers relative to unimpaired APOE4 carriers (Fig. 2a–d). Additionally, global/memory impaired APOE4 carriers also showed slightly decreased hippocampal volume relative to impaired APOE3 homozygotes (p = 0.02), and unimpaired APOE4 carriers had slightly greater posterior cingulate volume than unimpaired APOE3 homozygotes (p = 0.02); however, these smaller effects did not survive FDR correction for multiple comparisons. Mediation analyses confirmed effects of hippocampal (p = 0.0007) and parahippocampal (p = 0.01) volume reduction on global/memory impairment, independent of regional CBF (Fig. 2e–f). In the posterior cingulate, regional CBF had a direct effect on global/memory impairment independent of volume reduction (p = 0.02), and volume reduction showed a nonsignificant trend (p = 0.05) towards an effect on global/memory impairment independent of regional CBF (Fig. 2g). No significant indirect mediation effects were observed. All models controlled for age, sex, education, and VRF burden.

Fig. 2.

a Volumetric ROIs are displayed for regions implicated in global/memory impairment, including hippocampus (HC), parahippocampal gyrus (PHG), and posterior cingulate gyrus (PCG). b–d Regional volumes are shown in comparisons of APOE4 status (APOE3 homozygotes [blue] vs. APOE4 carriers [red]) by global/memory impairment by clinical dementia rating (CDR) global score (0 vs. 0.5): CDR 0 APOE3 (n = 70), CDR 0 APOE4 (n = 53), CDR 0.5 APOE3 (n = 19), CDR 0.5 APOE4 (n = 24). e–g Results of mediation analysis are displayed for APOE4 carriers. Purple line indicates significant direct effect of CBF on global/memory impairment in PCG independent of regional volume, and green lines indicate significant effects of volume reduction on global/memory impairment in HC and PHG, independent of regional CBF. All analyses controlled for age, sex, education, and VRF burden

Structural integrity of the uncinate fasciculus white matter tract connected to hippocampal and parahippocamal global/memory regions showed decreased integrity (DTI FA) in global/memory impaired APOE4 carriers relative to unimpaired APOE4 carriers (p = 0.003), and segment-level analysis showed significant differences in uncinate fasciculus segments more proximal to hippocampal and parahippocampal regions (Fig. 3a–d). Mediation analysis confirmed the effect of uncinate fasciculus integrity on global/memory impairment in APOE4 carriers, independent of regional CBF in the hippocampus (p = 0.02) and parahippocampal gyrus (p = 0.02) (Fig. e–f). Structural integrity of the cingulum white matter tract connected to the posterior cingulate global/memory region showed decreased integrity in global/memory impaired APOE4 carriers relative to unimpaired APOE4 carriers (p = 0.0004), and segment-level analysis showed significant differences in cingulum segments more proximal to the posterior cingulate and hippocampal and parahippocampal regions (Fig. 3g–i). Mediation analysis revealed that the effect of posterior cingulate CBF on global/memory impairment in APOE4 carriers was not direct but was mediated by decreased white matter integrity in the connecting cingulum tract, indirect effect β = − 0.7936, 95% CI (− 124.1293, − 0.1541) (Fig. 3g–j). All models controlled for age, sex, education, and VRF burden.

Fig. 3.

a White matter tract fractional anisotropy (FA) maps of the uncinate fasciculus (UF) and cingulum tracts connected to grey matter regions supporting global/memory functions. b Examples of individual UF tracts for each group. c UF tract segment-level differences in (left and right) UF FA between groups are displayed with red segments representing areas with significantly (p <.05) lower FA in CDR 0.5 vs. 0 by APOE carrier status. d UF FA values averaged over the entire tract and across left and right tracts are displayed in violin plots with median and interquartile range and compared by APOE status (APOE3 homozygotes [blue] vs. APOE4 carriers [red]) and global/memory impairment by clinical dementia rating (CDR) global score (0 vs. 0.5): CDR 0 APOE3 (n = 41), CDR 0 APOE4 (n = 34), and for CDR 0.5 APOE3 (n = 8), CDR 0.5 APOE4 (n = 11). e–f Mediation analysis for APOE4 carriers with green lines indicating significant effects of UF FA, independent of CBF in connected hippocampal (HC) and parahippcampal (PHC) grey matter. g Examples of individual cingulum tracts. h Segment-level differences in cingulum FA. i Cingulum FA values compared by group: CDR 0 APOE3 (n = 42) APOE4 (n = 34), CDR 0.5 APOE3 (n = 8) APOE4 (n = 11). j Mediation analysis for APOE4 carriers with red lines indicating posterior cingulate gyrus (PCG) CBF effects mediated by Cingulum FA. All analyses controlled for age, sex, education, and VRF burden

Exploratory analysis of CBF and structural integrity in regions underlying impairment in attention/executive and language domains identified effects in the combined sample of APOE4 carriers and APOE3 homozygotes. First, we analyzed brain structures underpinning attention/executive abilities, including frontal and inferior parietal CBF and structural volume, and white matter structural integrity in the superior longitudinal fasciculus tract that connects frontal and inferior parietal regions. Participants with attention/executive impairment in the combined sample exhibited decreased frontal CBF (p = 0.04) and a trend (p = 0.05) towards a significantly decreased inferior parietal CBF relative to those who were unimpaired (Fig. 4a–d). Structural analysis indicated no differences in grey matter volume within these structures (data not shown), but both APOE4 carriers and APOE3 homozygotes with attention/executive impairment exhibited decreased white matter integrity in the superior longitudinal fasciculus tract that connects frontal and parietal regions relative to APOE4 carriers and APOE3 homozygotes who were unimpaired, and segment level analysis showed varied distribution of significant differences along the tract (Fig. 4e–h). Mediation analysis confirmed direct effects of frontal CBF (p = 0.02) and inferior parietal CBF (p = 0.04) on attention/executive impairment, independent of superior longitudinal fasciculus integrity. In addition to these independent effects of CBF, there were effects of frontal CBF on superior longitudinal fasciculus integrity (p = 0.04), effects of superior longitudinal fasciculus integrity on attention/executive impairment (p = 0.02), effects of inferior parietal CBF on superior longitudinal fasciculus integrity (p = 0.01), and effects of superior longitudinal fasciculus integrity on attention/executive impairment, p = 0.02 (Fig. 4i–j). However, neither of the indirect effects from the mediation analyses were statistically significant: frontal CBF, β = − 0.4458, 95% CI (− 1.7433, 0.0615); inferior parietal CBF, β = − 0.4345, 95% CI (− 1.7954, 0.0282).

Fig. 4.

a Representative frontal lobe (superior, middle, and inferior frontal gyri) CBF map. Scale bar is − 1 to + 2 residualized z score. b Regional frontal lobe CBF compared by group APOE status (APOE3 homozygotes [blue] vs. APOE4 carriers [red]) and attention/executive impairment: unimpaired APOE3 (n = 51), unimpaired APOE4 (n = 44), impaired APOE3 (n = 14), impaired APOE4 (n = 9). c Representative inferior parietal lobe (IPL) CBF map. Scale bar is − 1 to + 2 residualized z score. d Regional frontal lobe CBF values compared by group: unimpaired APOE3 (n = 51), unimpaired APOE4 (n = 44), impaired APOE3 (n = 14), impaired APOE4 (n = 9). e Fractional anisotropy (FA) maps of the superior longitudinal fasciculus (SLF) connecting frontal and IPL grey matter regions supporting attention/executive functions. f Examples of individual SLF tracts. g Segment-level tract (left and right) differences in SLF FA between groups are shown in red. h SLF FA values averaged (left and right) along the entire tract and compared by group: unimpaired APOE3 (n = 28), unimpaired APOE4 (n = 26), impaired APOE3 (n = 5), unimpaired APOE4 (n = 6). i Mediation analysis for combined sample of APOE3 homozygotes and APOE4 carriers with purple lines indicating significant direct effects of CBF, independent of SLF FA, and red lines indicate significant CBF effects on SLF FA and SLF FA effects on attention/executive impairment. j Mediation analysis for combined sample of APOE3 homozygotes and APOE4 carriers with purple lines indicating significant direct effects of CBF, independent of SLF FA, and red lines indicate significant CBF effects on SLF FA and SLF FA effects on attention/executive impairment. All analyses controlled for age, sex, education, and VRF burden

Finally, we analyzed regional CBF and structural volume in left inferior frontal regions involved in language ability. Participants with language impairment in the overall sample of combined APOE4 carriers and APOE3 homozygotes exhibited a trend towards decreased left inferior frontal CBF (p = 0.047), and both APOE4 carriers (p = 0.003) and APOE3 homozygotes (p = 0.04) exhibited decreased grey matter volume, relative to those who were unimpaired (Fig. 5a–d). Mediation analysis confirmed direct effects of left inferior frontal CBF on language impairment independent of inferior frontal volume (p = 0.04), as well as effects of inferior frontal volume independent of inferior frontal CBF, p = 0.007 (Fig. 5d).

Fig. 5.

a Representative left inferior frontal lobe CBF map. Scale bar is − 1 to + 2 residualized z score. b Regional left inferior frontal lobe CBF values and c volumes compared by APOE status (APOE3 homozygotes [blue] vs. APOE4 carriers [red]) and language impairment: unimpaired APOE3 (n = 52), unimpaired APOE4 (n = 45), impaired APOE3 (n = 15), impaired APOE4 (n = 14). d Mediation analysis for the combined sample of APOE3 and APOE4 carriers with purple line indicating significant direct effect of CBF, independent of regional volume, and green line indicating significant effect of volume reduction, independent of regional CBF. All models controlled for age, sex, education, and VRF burden

Discussion

Here we report a pattern of regional CBF and structural differences in cognitively impaired relative to unimpaired APOE4 carriers and APOE3 homozygotes, with both independent and structurally mediated CBF effects on different cognitive ability domains. The findings overall suggest that APOE4 carriers display a distinct pattern of regional CBF that is characterized by posterior cingulate gyrus hypoperfusion that is related to impairment in memory and global cognition through decreased cingulum integrity. These findings are consistent with prior studies implicating posterior cingulate hypoperfusion in cognitive impairment [51, 52], and reveal an apparent susceptibility of APOE4 carriers to hypoperfusion in this region. Findings further suggest that the contribution of posterior cingulate hypoperfusion to cognitive impairment may be related to injury of the connected cingulum white matter, particularly in areas proximal to connections of the posterior cingulate gyrus and medial temporal lobe. Together the findings are consistent with the established susceptibility of white matter to hypoperfusion [53], the importance of white matter injury in vascular contributions to cognitive impairment [54], and preclinical studies demonstrating APOE4 causes white matter injury and cognitive impairment through neurovascular dysfunction [13]. These findings have major clinical significance given APOE4 carriers with global/memory impairment are at very high risk for progression to dementia [55].

We did not find any differences in hippocampal or parahippocampal CBF in relation to APOE4 status or cognitive status, but we did find an expected decrease in the volume of these brain structures in APOE4 carriers with global/memory impairment. Importantly, these volume differences were found to have statistical effects on global/memory impairment that were independent of regional CBF, potentially suggesting that neurodegeneration within the hippocampus and parahippocampal gyrus in APOE4 carriers is not driven by hypoperfusion injury. This is consistent with prior perfusion imaging studies focusing on these memory regions in early-stage cognitive impairment, which have shown mixed results, including increased, decreased, and no changes relative to unimpaired controls [16, 21, 25, 51, 52]. We also observed both decreased CBF and decreased volume in the posterior cingulate in APOE4 carriers specifically, but the statistical effects of posterior cingulate CBF and volume on global/memory impairment were independent. This could suggest distinct effects of hypoperfusion and neurodegeneration on global/memory impairment in APOE4 carriers.

The dominant interpretation of posterior cingulate hypoperfusion in older adults with cognitive impairment is that it represents secondary effects of neurodegeneration and decreased neuronal metabolism [56]. This hypothesis is supported by studies showing strong correlations between perfusion MRI and fluorodeoxyglucose-positron emission tomography (FDG-PET) [57]. However, it has also been proposed that the hypoperfusion of this region could be due to vascular dysfunction and may contribute to cognitive decline [58]. This alternative vascular hypothesis is supported by reports of cerebrovascular reactivity deficits in the posterior cingulate gyrus in older adults with vascular disease [59]. Further support for the vascular hypothesis comes from studies showing that cerebral hypoperfusion predicts cognitive decline and dementia independent of neuronal metabolic deficits measured by FDG-PET [6, 8]. The present findings offer fresh insight into this debate since posterior cingulate CBF and volume differences showed independent statistical effects on global/memory impairment, potentially suggesting distinct vascular and neurodegenerative effects on cognitive impairment in this region. The fact that cingulum white matter integrity mediated the effects of posterior cingulate hypoperfusion on global/memory impairment further supports the concept that hypoperfusion in this region may represent a vascular mechanism of brain injury.

Findings from additional exploratory analyses in both APOE4 carriers and APOE3 homozygotes suggest a pattern of frontal and inferior parietal hypoperfusion that contributes to deficits in attention and executive function both directly, independent of structural changes, and through decreased integrity of the superior longitudinal fasciculus white matter tract that connects these two regions. However, indirect mediation effects in these regions did not reach statistical significance, preventing a definitive conclusion from being drawn regarding independent versus mediating effects of CBF and white matter integrity on attention/executive impairment. Findings in the left inferior frontal gyrus also suggest hypoperfusion in this region may contribute to language impairment, independent of regional volume. Thus, hypoperfusion in other cortical regions outside the posterior cingulate may be less specific to APOE4, and could contribute to non-amnestic cognitive impairment either by directly impairing neurocognitive function or through its relationship with white matter injury in connected tracts.

We observed slightly increased CBF and regional grey matter volume within the posterior cingulate cortex in cognitively unimpaired APOE4 carriers relative to cognitively unimpaired APOE3/3 homozygotes. This contrasts with the APOE4-specific decreases in CBF and volume observed within the posterior cingulate region in those with cognitive impairment. However, these differences did not survive multiple comparison correction and are of unclear clinical significance. It is possible slight increases in CBF and grey matter volume in this region in cognitively unimpaired APOE4 carriers represent early pathologic or compensatory changes related to genetic risk for cognitive decline. It is also possible that higher baseline CBF and volume in this susceptibility region may help to maintain normal cognition despite the presence of APOE4. Further insight into these questions may be gleaned from longitudinal and predictive studies examining whether the observed increases in posterior cingulate CBF and volume are longstanding or represent compensatory or pathological increases that may differentially predict future cognitive decline.

The strengths of the present study include the use of multimodal MRI to investigate regional CBF and grey and white matter structural integrity in APOE4 carriers and APOE3 homozygotes with and without cognitive impairment in three cognitive ability domains, as well as the mediation analysis identifying direct, independent, and indirectly mediated effects of regional CBF on cognitive impairment. Study limitations include the cross-sectional design, limiting causal inference, and the inability to study white matter hypoperfusion directly with pCASL-MRI. The sample size was also relatively small, limiting statistical power. Further longitudinal mediation studies may clarify the relative timing and predictive value of changes in regional CBF and corresponding structural integrity, yielding additional insights into the independent and interactive effects of vascular and neurodegenerative processes driving cognitive decline in APOE4 carriers at genetic risk for dementia.

Author contribution

Study design and conceptualization: IP, SB, AC, BVZ, DAN. Data acquisition: IP, EBJ, JMR, LSS, HCC, AWT, BVZ. Imaging data processing and analysis: IP, TL, SD, AK, ACE, JPMA, DAN. Statistical analysis: IP, TL, AK, MHCL, DAN. Manuscript preparation: All authors.

Funding

Open access funding provided by SCELC, Statewide California Electronic Library Consortium. This research was supported by National Institutes of Health grants P01 AG052350, AT, BVZ, DAN; P30 AG066530, HC; (DAN: R01 AG064228, R01 AG060049, R01 AG082073), and the Canadian Institutes of Health Research (AK: DFD- 170763).

Data availability

The anonymous data that support the findings of this study are available upon reasonable request from the corresponding author, DAN, through appropriate data sharing protocols.

Declarations

Consent for publication

This manuscript was reviewed and approved by an internal committee to ensure scientific integrity prior to publication. All images were analyzed using Imagetwin software to confirm no images were improperly altered or duplicated.

Competing interests

The authors declare no competing interests.