Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline and memory loss. The APOE ε4 allele, the strongest genetic risk factor for AD, plays a key role in lipid metabolism, inflammatory response, and neurovascular function (Caselli et al., 2009; Jackson et al., 2021; Peng et al., 2018; Tzioras et al., 2019). Furthermore, APOE ε4 has been associated with faster rates of progression in AD (Kanai et al., 1999; Freudenberg-Hua et al., 2018). It may also act as a modifier of cognitive decline through its interaction with cardiovascular diseases (CVDs) (Scuteri et al., 2005; Wei et al., 2024). The combination of APOE ε4 dosage and CVDs may work synergistically to worsen cognitive outcomes in individuals with both increased genetic and vascular risks (Lee B. C. et al., 2023; Levine et al., 2023).

AD is characterized by early and progressive loss of cholinergic neurons in the NBM, a key nucleus of the basal forebrain (Freudenberg-Hua et al., 2018; Rogers et al., 1985). Furthermore, studies have shown that cholinergic neuron loss in NBM is an important feature of the pathogenesis of neuritic plaque counts, with the APOE ε4 allele specifically impairing cholinergic function in animal and neuropathological studies, independent of plaque presence (Arendt et al., 1985; Allen et al., 1997). Conversely, a pathological study has shown sparing of NBM neurons in cases with vascular dementia (Jung et al., 2012). Early detection of in vivo changes in the NBM in APOE ε4 carriers may provide an understanding of the downstream interaction effects of APOE ε4 and cardiovascular risk in aging adults.

The basal forebrain is among the earliest brain regions affected by aging and vascular pathology, even in the absence of overt Alzheimer’s disease. NBM, a specific subregion in the basal forebrain, receives input from the amygdala and hippocampus and also projects to widespread cortical regions (Mihailoff et al., 1989; George et al., 2011; Zaborszky et al., 2015). Notably, the NBM demonstrates selective vulnerability to both cholinergic degeneration and cerebrovascular burden. Subcortical networks involving the NBM may be important in understanding the progression of AD (Shu et al., 2019). Advanced dMRI techniques could aid in elucidating the progressive nature of AD and the neurobiological effect of APOE ε4.

Free-water (FW) imaging has been shown to be sensitive to detecting in vivo regional microstructural changes in AD (Ofori et al., 2019; Ofori et al., 2024; Archer et al., 2021). The majority of these studies have focused on cortical changes in white matter, and mediotemporal gray matter FW changes have shown promise for detecting changes in the neurodegenerative process (Chu et al., 2022; Maillard et al., 2019; Edde et al., 2020; Montal et al., 2018). FW molecules do not experience bulk flow and are minimally restricted by cellular or extracellular barriers; free water is prone to accumulate in the extracellular space that arises from neuronal injury and neuroinflammatory processes. (Pasternak et al., 2012; Ofori et al., 2015). Early detection may result in a decrease in FW, which would reflect increased cellularity and/or glial recruitment, whereas increases in FW may reflect cellular loss and increased osmotic challenges in the cell (Pasternak et al., 2014; Starck et al., 2021). Most studies in AD populations have found no effect of APOE ε4 as a covariate on the FW signal in these regions (Ofori et al., 2019; Chu et al., 2022).

Amyloid-beta (Aβ) and tau pathology have been shown to be elevated in the basal forebrain of AD patients, relative to controls (Mieling et al., 2023; Nakaya et al., 2022). Interestingly, pathology studies have shown an inverse relationship between choline acetyltransferase activity in the temporal cortex with increasing APOE ε4 allele copy and an increase in senile plaques and tangles in cortical areas (Beffert and Poirier, 1996). As therapies have been developed to slow the causal influence of AD pathological hallmarks, it is apparent that understanding cholinergic system integrity along with AD pathology may be important for prodromal AD (Ashford, 2015).

Recent inquiries have further delineated the APOE ε4 role in modulating inflammatory pathways within the neurodegenerative context (Mieling et al., 2023; Reiman et al., 2024). The APOE ε4 gene dose-related brain imaging findings have shown increased amyloid, microglial activation, and translocator protein (TSPO) positron emission tomography (PET) (Reiman et al., 2009; Ferrari-Souza et al., 2023; Fan et al., 2015). Immune modulation in the NBM may occur in response to amyloid-beta or tau pathology; however, it is likely subdetectable with the resolution of most PET studies and may instead be detectable using FW imaging.

APOE ε4 impacts cholesterol metabolism, which may influence amyloid-beta and vascular pathogenesis. A recent report suggested that blood-based cardiovascular dysfunction markers are related to brain FW (Mayer et al., 2022; Lee H. et al., 2023); however, it did not examine the effects of APOE ε4. The APOE genotype influences blood vessels throughout life, beginning in early development. For instance, an autopsy study found differences in peripheral LDL cholesterol and atherosclerotic plaque formation in vascular structures following a clear pattern, with APOE ε4 carriers having the greatest burden of levels (Hixson, 1991). Accordingly, several reports have indicated that APOE ε4 carriers have increased odds for cardiovascular diseases (Lahoz et al., 2001; Jofre-Monseny et al., 2008; Weiss et al., 2021).

Collectively, our study aims to investigate the effects of APOE ε4 and cardiovascular risk on free-water metrics of the NBM and evaluate relationships with cardiovascular measurements. We hypothesize that homozygous carriers of the APOE ε4 gene will have increased FW in the NBM, and this may reflect early neurodegeneration, as shown by studies indicating differences between early mild cognitive impairment (MCI) and cognitively unimpaired. We also suggest that increases in FW would relate to cardiovascular risk measurements in APOE ε4 carriers.

The present investigation sourced its biological specimens and associated data from the Arizona APOE Cohort, a longitudinal endeavor tracking cognitive health across individuals bearing varying quantities of the APOE ε4 allele, ranging from none to two copies between 1994 and 2017 (Reiman et al., 2009; Fan et al., 2015). Prior to their participation, all volunteers provided their written informed consent, underscoring the ethical rigor of the study, which received the endorsement of the institutional review boards of both the Mayo Clinic and Banner Health in the surrounding areas of Phoenix.

Evaluative protocols within the study encompassed a comprehensive neurological assessment, alongside a suite of diagnostic tools, including the Folstein Mini-Mental Status Exam (MMSE) for cognitive function (Folstein et al., 1975), and the Hamilton Depression Rating Scale for mood assessment (Hamilton, 1960), Instrumental Activities of Daily Living for evaluating the participants’ ability to perform daily tasks. Furthermore, the Structured Psychiatric Interview for DSM-III-R was used to scrutinize psychiatric conditions, supplemented by a diverse array of neuropsychological examinations and additional clinical evaluations. Importantly, at the inception of the study, none of the participants fulfilled the established diagnostic criteria for amnestic mild cognitive impairment (aMCI), AD, any other dementia variants, or major depressive disorder, ensuring the cognitive integrity and psychological health of the cohort at baseline.

The current analysis included 167 subjects with a mean age of 64.0 ± 7.1 years (range: 49–81) and a mean education level of 16.5 ± 2.0 years. Seventy-five percent (75%) of subjects were women, and 15% were Hispanic. The sample was derived from the initial collection of diffusion imaging data; however, the timing of this collection varied among individuals.

Imaging data from MRI were processed utilizing a custom suite of tools, including the Functional MRI of the Brain (FMRIB) Software Library (FSL), Advanced Normalization Tools, and previously developed UNIX shell scripts (Ofori et al., 2019; Ofori et al., 2015). The pipeline for dMRI processing was fully automated. The correction of scans for distortions induced by eddy currents and head movement was accomplished through affine transformations, with a subsequent adjustment of gradient directions to account for these corrections. Additionally, the elimination of non-brain tissues from the scans was performed. FW is a quantifiable metric derived from dMRI scans, indicating the fractional volume of water that diffuses freely in each voxel. FW maps were generated using a specialized MATLAB script, and the approach adopted for this task aligns with methodologies documented in previous studies (Pasternak et al., 2012; Ofori et al., 2015). Quality control checks were conducted with an eddy current outlier report, visual inspection, and correction for susceptibility. Any motion exceeding 2 mm on average movement, 5 mm on maximal movement, and having > 2% of slice outliers was removed from analysis. The average movement was ~0.9 mm per patient. The spatial distribution of FW across seven predetermined regions (i.e., left and right posterior NBM, left and right NBM, left and right transentorhinal tract to locus coeruleus, and bilateral anterior nigra) of interest (ROIs) is illustrated in Figure 1. Furthermore, diffusion tensors adjusted to negate the partial volume effects of FW were generated, from which fractional anisotropy was calculated using FSL’s DTIFit, resulting in the creation of FW-corrected fractional anisotropy (FA_c) maps.

The maps for FW in subject space were aligned to the Montreal Neurological Institute (MNI) 152 template space. We computed FA maps in subject space and then warped FA maps to MNI space, 1 mm isotropic FMRIB58 FA template (Jenkinson et al., 2012). An initial 12-d.o.f. affine transform (mutual-information cost) was followed by a non-linear SyN warp. The resulting deformation field was applied to all diffusion-derived maps. To quantify spatial fidelity, we first binarized each subject’s warped FA map and then we then compared this mask with the 50%-thresholded probabilistic NBM from Zaborszky et al. (2008) using ANTs ImageMath DiceAndMinDistSum. Across 167 participants, the mean ± SD Dice similarity was 0.88 ± 0.04, and the 95th-percentile Hausdorff distance was 0.43 ± 0.11 mm.

We sought to investigate whether free-water imaging could detect differences in the NBM across varying APOE genotypes. It has been well established that neuronal loss within the cholinergic NBM cascades events that lead to mediotemporal lobe degeneration (Ofori et al., 2015). Our findings indicated decreased NBM FW values in APOE ε4/4 carriers when compared to APOE ε3/4 and non-carriers. We also found significant associations of increased NBM FW values with increased homocysteine levels and systolic BP for APOE ε3/3 carriers. In APOE ε4/4 carriers, a unique association of increased NBM FW values with decreased cholesterol to HDL ratios indicates that APOE may interact with NBM to modulate lipid metabolism. We posit that decreased NBM FW in APOE ε4/4 may reflect that APOE ε4/4 increases cellularity in the NBM through altered neurovascular and neuroinflammatory processes in comparison to non-carriers of APOE ε4. Our findings have various implications for how FW imaging may be useful for aging, genetic influences on the brain, and AD research.

Previous studies have indicated that APOE ε4 exacerbates neuronal loss in the NBM, while other studies have not, as this may be due to differences in the ability to quantify neuronal loss (Zhao et al., 2020; Lai et al., 2021). Furthermore, the association between APOE ε4 and neurodegeneration may be due to its accelerated effects through either hypometabolism, amyloid, or tau (Lerch et al., 2022; Snellman et al., 2023). Our findings are in line with others that APOE ε4 carriers show altered brain characteristics compared to non-carriers (Reiman et al., 2009). We did not observe FW changes in the LC to the entorhinal cortex tracts, where tau pathology is suggested to begin, and would result in FW changes that have been seen in other reports (Chu et al., 2022). One difference between the current report and previous investigations is the cognitive status of participants. Other studies have compared cross-sectional data across impaired vs. intact cognitive status, when presumably the pathological process has been occurring (Ofori et al., 2019). Our cohort has no cognitive impairment and a cohort mean age of 64 years vs. 71 years or older in previous studies. Furthermore, previous studies covaried for APOE ε4 or combined heterozygous and homozygous carriers in the APOE ε4 group. We were able to evaluate dose effects in regions potentially susceptible to AD and found differences only in the NBM.

Our findings are in line with previous findings suggesting that FW may reflect downstream consequences of hemodynamic alterations (Rydhög et al., 2017; Edwards et al., 2024). We found that higher NBM FW levels were associated with increases in systolic BP in non-carriers and APOE ε3/4. A previous report from the Framingham Heart study in over 2000 adults with a mean age of ~55 showed that systolic BP has a direct influence on cerebral FW white matter, and this is moderated by carotid pulse wave velocity (Maillard et al., 2017). We extend these findings to suggest that hemodynamic alterations may also influence gray matter FW values in APOE ε4 non-carriers. We did not find a significant association between NBM FW and systolic BP in APOE ε4/4 carriers. It has been suggested that APOE ε4 and BP interact to increase the risk of small vessel disease (Mayer et al., 2022; Walker et al., 2017; Biffi et al., 2019). APOE plays a critical role in determining the risk and severity of cerebral amyloid angiopathy, which is characterized by the accumulation of amyloid-beta. One report found a correlation between white matter FW and diastolic BP when controlling for various demographic and lifestyle variables in individuals with metabolic syndrome (Andica et al., 2023). The lack of correlation in only APOE ε4 carriers may be due to a multitude of factors, and APOE ε3 may play a key role in modifying BP through the interaction with the hypothalamus (Carmichael and Wainford, 2015).

The novel finding of increased homocysteine levels positively related to increased NBM FW levels across APOE ε4/4 and non-carriers suggests that increased free-water levels may reflect neurovascular pathology. Elevated homocysteine levels, a risk factor for cardiovascular disease, have been observed in individuals carrying the APOE ε4 allele. Homocysteine is known to contribute to endothelial dysfunction and atherosclerosis. Our findings are similar to another study that showed increased cerebral WM FW was positively associated with increased cardiovascular health markers (Ji et al., 2023). One difference between the current study and Ji et al. is that the association was found in our cognitively normal participants, whereas significant correlations in the Ji et al. study were only found in individuals experiencing some form of cognitive impairment (Ji et al., 2023). Our findings are also in line with findings showing that elevated homocysteine is associated with increased atrophy in the medial temporal lobe in aMCI (Ma et al., 2022). There is growing evidence that homocysteine levels are related to cognitive impairment, as individuals with AD have greater vitamin B deficiencies. Our findings suggest that NBM FW values and homocysteine may provide insight into the longitudinal effects on cognition.

Decreased NBM FW could reflect increased lipid metabolism through increasing extracellular cholesterol in APOE ε4 carriers. Cholesterol can control the levels and dynamics of cholinergic neurons and modulates their diffusion. Increased brain cholesterol has been suggested to occur due to amyloid buildup (Piccarducci et al., 2023). Our findings of decreased extracellular space levels with increased cholesterol to HDL ratios are in line with this potential mechanism. It has been shown that myelin breakdown releases excess cholesterol into the extracellular space. One report suggests that plasma 24-hydroxycholesterol (24-OHC) may be a sensitive marker of increased cholesterol metabolism during acute demyelination early in the neurodegenerative processes prior to neuron loss (Hughes et al., 2012). Although we do not have measures of brain cholesterol, perhaps future studies examining the interconnectedness between brain and peripheral lipid metabolism and free water may help unravel how cholesterol metabolism is affecting NBM FW.

The exact pathophysiological mechanism behind the accumulation of brain extracellular water remains unclear. One of the main advantages of diffusion imaging is that it provides microstructural information that can be related to processes within the tissue. Current research has not definitively determined whether the increase in extracellular water is due to neurodegenerative processes, vascular pathology, or a combination of both. In the context of neurodegenerative diseases, increases in FW may reflect microstructural damage and increased CSF in the region, whereas decreases in FW may reflect gliosis or cellularity (Dubelaar et al., 2004). Recent studies have demonstrated that amyloid-beta interacts with cortical tau pathology to influence neurodegeneration. Aβ toxicity is expected to cause inflammatory changes detectable through FW imaging. Furthermore, FW has been found to be correlated with plasma Aβ42/40 in PET-negative participants without dementia, although these patients had altered cognitive status (DeSimone et al., 2024). Our findings showed a decrease in cognitively normal APOE ε4 carriers. This may reflect a critical window prior to tau involvement that could impact NBM microstructural environments. Interestingly, FA differences were not uniformly lower in higher cardiovascular risk groups, suggesting that FA may reflect both axonal degeneration and compensatory remodeling depending on APOE genotype and regional vulnerability. Our NBM findings are synergistic with the lack of FW or FA_c differences in the locus coeruleus to entorhinal tracts. Given the accumulation of studies showing elevated FW changes when presumably the pathophysiological processes have occurred, our findings may suggest that increases in FW reflect potential influences of tau accumulation, and where FW decreases may reflect increased glial cellularity. The increase in glial cells is likely part of a neuroinflammatory response, as it has been shown that APOE ε4/4 may recruit these cells at an enhanced rate. An increase in the number and size of glial cells would decrease the extracellular volume (that is, decreased FW measure). Interestingly, there are few studies examining the effects of disease-related genetic factors on free-water imaging. Further studies would be required to determine whether these differences are related to long-term symptom changes.

We identified a left-lateralized association between HDL-to-cholesterol ratios and NBM fractional water. Previous tracing and neuroimaging studies have demonstrated that the NBM exhibits denser projections to limbic and association cortices in the left hemisphere (Zaborszky et al., 2015). This anatomical asymmetry may potentially explain the observed left-hemispheric sensitivity to lipid dysregulation. Furthermore, a recent [¹¹C]PiB PET study, cognitively normal APOE ε4/4 carriers showed greater left-than-right cortical amyloid burden (Kjeldsen et al., 2022). Most reports typically average the right and left regions. The study also indicated that APOE ε4 does not appear to bias the spatial distribution of pathology toward each hemisphere (Kjeldsen et al., 2022). Additional research is necessary to fully understand the mechanisms underlying lateralized neurodegeneration in individuals with the APOE ε4.

Importantly, the use of single-shell imaging to compute free-water has been criticized, given its constrained ability to tease apart partial volume effects from true microstructural alterations. Nonetheless, our observation of a measurable FW decrease in cognitively normal APOE ε4 carriers shows that single-shell models can still detect physiologically relevant variations. The primary benefit of single-shell approaches is their feasibility in retrospective or multi-site studies, where the availability of multi-shell data may be limited. However, because single-shell methods may conflate different diffusion compartments, the precision of FW estimations can be adversely impacted and tissue measures underestimated. Replicating these findings with multi-shell acquisitions or complementary techniques would provide more rigorous confirmation of the underlying neurobiological events. More studies also showing the adverse impact of FW and corrected measures diffusion metrics on gray matter and vascular-rich areas are needed.

Our study has some limitations, as there was a small number of APOE ε4/4 carriers that had peripheral measurements compared to APOE ε3/4 and non-carriers, limiting the statistical power of our correlations. Furthermore, we did not have inclusion of APOE ε2 carriers, which may also limit the generalizability of our findings. APOE ε2 has been associated with potential neuroprotective effects, including reduced amyloid burden and possibly altered inflammatory responses, which could modulate FW measurements. The lack of full PET characterization in all subjects limits the generalizability of all APOE ε4/4 carriers. Initial diffusion metrics were collected at differing times throughout enrollment, and this may have some influence on the findings we observed. Lack of ethnic and racial diversity, over-representation of women, lack of integrated information on menopause status or hormonal use, and a cross-sectional approach that limits insights into the onset of differences between APOE groups. Future investigations should aim to link NBM FW, cardiovascular, and inflammatory marker values with cognitive aging trajectories to determine the functional influences of APOE-related allelic differences.

This is the first investigation to characterize the effect of APOE ε4 on FW imaging in cognitively unimpaired late middle-aged older adults. We reveal that APOE ε4/4 carriers have decreased NBM FW levels in comparison to APOE ε3/4 and non-carriers, whereas no differences were found in the locus coeruleus to entorhinal track. APOE ε4/4 also had an inverse relation between increased free-water and an elevated cholesterol/HDL ratio, perhaps suggesting that elevated cholesterol levels impinge on the NBM extracellular space. Furthermore, NBM FW levels were related to peripheral markers of homocysteine and systolic BP, suggesting that FW may also reflect downstream biomechanical and chemical modulators of vascular dysfunction. These differences may reflect modulatory inflammatory processes in APOE ε4 carriers, as FW techniques are optimal to estimate neuroinflammation with MRI. The involvement of the APOE ε4 allele in modulating these processes, particularly its impact on the NBM and its potential role in AD pathogenesis, warrants further investigation. Elucidating the complex interplay between genetic, vascular, and inflammatory factors may provide valuable insights into the mechanisms underlying cognitive impairment and neurodegenerative diseases, ultimately informing the development of targeted therapeutic strategies.