Apolipoprotein E4 moderates the association between vascular risk factors & brain pathology

Abstract

Background:

The strongest genetic risk factor for late-onset Alzheimer’s disease (AD), Apolipoprotein E4 (APOE4), increases cardiovascular disease risk and may also act synergistically with vascular risk factors to contribute to AD pathogenesis. Here, we assess the interaction between APOE4 and vascular risk on cerebrovascular dysfunction and brain pathology.

Methods:

This is an observational study of cognitively normal older adults, which included PET imaging and vascular risk factors. We measured beat-to-beat blood pressure (MAP) and middle cerebral artery velocity (MCAv) at rest and during moderate-intensity exercise. Cerebrovascular measures included cerebrovascular conductance index (CVCi) and the cerebrovascular response to exercise (ΔMCAv).

Results:

There was a significant interaction between resting CVCi and APOE4 carrier status on β-amyloid deposition (p=0.026), with poor conductance in the cerebrovasculature associated with elevated β-amyloid for the APOE4 carriers only. There was a significant interaction between non-high-density lipoprotein (non-HDL) cholesterol and APOE4 carrier status (p=0.014), with elevated non-HDL cholesterol predicting a blunted ΔMCAv with exercise in APOE4 carriers and the opposite relationship in non-carriers.

Conclusions:

Both cerebral and peripheral vascular risk factors are preferentially associated with brain pathology in APOE4 carriers. These findings provide insight into pathogenic vascular risk mechanisms and target strategies to potentially delay AD onset.

Introduction

Cerebrovascular dysfunction is increasingly recognized as an early, causal agent in the pathogenesis of late-onset Alzheimer’s disease (AD).^1–5 One prominent model of AD now places brain vasculature abnormalities – a historically under-recognized mechanism of disease development - ahead of all other biomarker changes, including β-amyloid and tau deposition.² Adults with the Apolipoprotein E4 (APOE4) allele, the strongest known genetic risk factor for AD, demonstrate an accelerated age-related cerebral blood flow (CBF) decline compared to non-carriers.⁶ Additionally, cognitively normal APOE4 carriers exhibit early blood brain barrier breakdown that occurs independently of β-amyloid and tau and predicts subsequent cognitive decline.³

The APOE4 allele not only increases risk of cardiovascular disease, such as myocardial infarction and stroke,⁷ but may also act synergistically with cardiovascular risk factors to promote dementia development.^8–13 For example, cognitive dysfunction is more significantly associated with vascular risk factors like high cholesterol in APOE4 carriers than non-carriers,⁹ suggesting elevated cholesterol (and particularly pro-atherogenic cholesterol)¹⁴ may preferentially promote brain pathology in APOE4 carriers. Additionally, APOE4 exacerbates the deleterious effects of cerebrovascular risk factors on neuropsychological performance,¹⁰ and stroke has been shown to act synergistically with the APOE4 allele to increase dementia risk.^{11, 13}

Cerebrovascular dysfunction has been posited to contribute causally to β-amyloid deposition in AD pathogenesis.¹⁵ Reduced CBF at rest occurs prior to β-amyloid deposition in APOE4 carriers.¹⁶ Subtle changes in cerebrovascular resistance and conductance can be detected even earlier than global CBF reductions in early stages of disease development.¹⁷ Therefore, a better understanding of the relationship between resting cerebrovascular conductance and β-amyloid load could provide insight into the connection between cerebrovascular perturbations and AD-associated brain pathology, particularly for APOE4 carriers. In addition to resting measures, assessing cerebrovascular function during exercise allows for a dynamic characterization of the cerebrovasculature. Changes in CBF during an acute bout of exercise reflect the cerebrovascular response to a wide variety of physiological inputs, including altered perfusion pressure, arterial blood gas, neural activity and brain metabolism.¹⁸ Our group previously demonstrated a blunted cerebrovascular response to moderate-intensity exercise with aging,¹⁹ stroke^{20, 21} and β-amyloid deposition.²² Importantly, our exercise stimulus provides a physiological challenge similar to common daily activities such as walking up a flight of stairs,²³ suggesting these perturbations could have implications for day-to-day functioning. Still, questions remain regarding the relationship between peripheral vascular risk factors and this dynamic response of the cerebrovasculature with acute exercise, particularly for APOE4 carriers who are at the highest known genetic risk of AD.

In the present analysis, we sought to investigate whether cerebrovascular dysfunction would more strongly predict β-amyloid load in APOE4 carriers than non-carriers. Additionally, we aimed to explore whether high levels of pro-atherogenic cholesterol, one of the strongest predictors of cardiovascular disease sequelae (including future stroke and myocardial infarction),²⁴ would be more strongly associated with cerebrovascular dysfunction in APOE4 carriers than non-carriers. To interrogate these aims, we assessed CBF velocity data collected with transcranial Doppler ultrasound (TCD) both at rest and during exercise in cognitively normal older adults (N = 54). We measured β-amyloid load by Positron Emission Tomography (PET)²² and quantified pro-atherogenic blood cholesterol as non-high-density lipoprotein (non-HDL) cholesterol.^{14, 24–26} We hypothesized that (1) poor resting cerebrovascular conductance (CVCi) would be more strongly associated with β-amyloid deposition in APOE4 carriers than non-carriers, (2) the acute cerebrovascular response to exercise (ΔMCAv) would be blunted in APOE4 carriers, and (3) this blunted ΔMCAv would be more strongly associated with elevated non-HDL cholesterol in APOE4 carriers than non-carriers.

Methods

Participants

As noted previously,^{22, 27–30} we recruited a convenience sample of older adults from a registry of individuals interested in research. Inclusion criteria were: (1) between 65 to 90 years of age, (2) cognitively normal/non-demented based on neuropsychological testing, (3) underactive or sedentary lifestyle, and (4) completion of a [18F-AV45] florbetapir positron emission tomography (PET) scan within six months of the vascular laboratory visit. Exclusion criteria were: diagnosis of diabetes, depression, congestive heart failure and inability to exercise. The University of Kansas Institutional Review Board approved all study procedures (IRB#: STUDY00001444). The study complied with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to data collection.

Positron Emission Tomography (PET) Scan for β-amyloid

Approximately 50 minutes after florbetapir (370 MBq) administration, the GE Discovery ST-16 PET/CT Scanner acquired two continuous 5-minute PET brain frames. We summed and attenuation corrected the frames. We then calculated standardized uptake value ratio to the whole cerebellum (SUVR) using a custom processing pipeline in SPM12 (http://www.fil.ion.ucl.ac.uk/spm), described previously.³⁰

Clinic visit

Medication information and cholesterol levels (total and high-density lipoprotein, HDL) were obtained during the clinic visit 1-2 months before the vascular laboratory visit. We chose to utilize non-HDL cholesterol (“bad” cholesterol)²⁵ as the predictor variable because studies have shown non-HDL cholesterol is an even more accurate marker for cardiovascular risk than LDL cholesterol.^{14, 24, 26} For example, one meta-analysis of 68 studies suggested non-HDL cholesterol was the best predictor for stroke and coronary artery disease among cholesterol measures.²⁴ This may be due to the fact that non-HDL cholesterol includes all pro-atherogenic particles in the blood, like very-low-density lipoprotein (VLDL) and intermediate-density lipoprotein (IDL) cholesterol, in addition to LDL.¹⁴ Non-HDL cholesterol was calculated as:

$$\text{Non-HDL cholesterol=Total cholesterol -HDL cholesterol}$$

³¹

Apolipoprotein E (APOE) genotyping

Frozen whole blood was assessed using a Taqman single nucleotide polymorphism (SNP) allelic discrimination assay (ThermoFisher) to determine APOE genotype. Taqman probes to the two APOE-defining SNPs, rs429358 (C_3084793_20) and rs7412 (C_904973_10), distinguished APOE4, APOE3, and APOE2 alleles. We excluded participants with an APOE2 allele (whether paired with APOE3 or APOE4), as this allele is associated with significantly lower risk of AD and can therefore confound the results when assessing for APOE4 effects. There were also no APOE4 homozygotes in our sample. Therefore, all participants were classified as either APOE4 carriers (APOE3/APOE4 genotype) or non-carriers (APOE3/APOE3 genotype).

Vascular laboratory setup

Visits to our laboratory occurred in the morning (beginning between 7:30 and 9 a.m.) for each participant and have been described in detail previously.^{22, 27–30} We monitored end-tidal carbon dioxide (P_ETCO₂) with a nasal cannula and capnograph (BCI Capnocheck 9004) and heart rate (HR) with a 5-lead electrocardiogram (Cardiocard, Nasiff Associates). We continuously recorded beat-to-beat blood pressure with a finger plethysmograph (Finometer Pro, Finapres Medical Systems), from which we calculated mean arterial pressure (MAP). We insonated the left middle cerebral artery velocity (MCAv) using a 2-MHz transcranial Doppler ultrasound (TCD) probe (RobotoC2MD, Multigon Industries) placed over the temporal window. TCD sonographers were blinded to β-amyloid status, APOE genotype, medication use and cholesterol profile.

Rest recording

The participant rested quietly for at least 15 minutes during setup. We then used an analog-to-digital data acquisition board (National Instruments) and custom-written script for MATLAB (v2015, Mathworks) to sample MCAv and MAP continuously at 500 Hz for 8 minutes of seated rest. MCAv and MAP were averaged over the 8 minutes, and the resting cerebrovascular conductance index (CVCi) was calculated as:

$$\text{CVCi=MCAv/MAP}$$

³²

Exercise recording

Participants were instructed to maintain a rate of 90 steps per minute on the recumbent stepper (NuStep T5XR). Resistance started at 40 Watts and was increased until each participant reached his or her pre-determined target HR range for moderate-intensity exercise (40-60% of age-predicted HR reserve). Once at target HR, data collection continued until the participant completed 8 minutes of continuous exercise in the moderate-intensity HR range. We averaged MCAv over the 8 minutes of exercise and calculated the cerebrovascular response to exercise (ΔMCAv) as:

$$\Delta \text{MCAv=mean MCAv during exercise-mean MCAv at rest}$$

²²

Statistical Analyses

We used SPSS Statistics (IBM) for all statistical analyses. We performed a multiple linear regression analysis to assess the effect of APOE4 carrier status, resting CVCi and the interaction between APOE4 carrier status and resting CVCi on β-amyloid deposition, controlling for age, sex, HDL cholesterol and statin use. Likewise, we performed a multiple linear regression analysis to assess the effect of APOE4 carrier status, non-HDL cholesterol and the interaction between non-HDL cholesterol and APOE4 carrier status on ΔMCAv, controlling for age, sex, β-amyloid deposition, HDL cholesterol and statin use. Finally, as an exploratory analysis to investigate the potential influence of brain volumes on these measures, we added gray and white matter volumes to each multiple regression model.

Results

We included 54 participants (65% female, 32% APOE4 carriers, 71.1 ± 5.5 years old) who had complete data for APOE genotype, β-amyloid load, cholesterol levels, and resting MCAv and MAP.

Cerebrovascular function at rest

We ran a multiple linear regression to predict β-amyloid deposition from age, sex, statin use, APOE4 carrier status, HDL cholesterol, non-HDL cholesterol, resting CVCi and the interaction between resting CVCi and APOE4 carrier status. There was linearity as assessed by partial regression plots and a plot of studentized residuals against the predicted values. There was independence of residuals, as assessed by a Durbin-Watson statistic near 2. There was no evidence of multicollinearity, as assessed by tolerance values greater than 0.1. There was homoscedasticity, as assessed by visual inspection of a plot of studentized residuals versus unstandardized predicted values. Data met the assumption of normality, as assessed by a P-P Plot. There were no outliers, as assessed by no standardized residuals greater than ±3 standard deviations (SD).

The multiple regression model significantly predicted β-amyloid deposition, F(8, 45) = 2.228, p = 0.043, R² = 0.284, adj. R² = 0.156. Regression coefficients and standard errors can be found in Table 1. There was a significant interaction between resting CVCi and APOE4 carrier status (p = 0.026), suggesting the relationship between CVCi and β-amyloid deposition is moderated by the APOE4 allele. Specifically, a larger resting CVCi, which denotes better cerebrovascular conductance, predicted lower β-amyloid deposition for APOE4 carriers but not non-carriers, while APOE4 carriers with poor conductance tended to have higher levels of β-amyloid deposition. There was no apparent relationship between CVCi and β-amyloid deposition for the non-carriers (see Fig. 1). There was no significant effect of APOE4 carrier status (p = 0.054) or resting CVCi (p = 0.260) on β-amyloid deposition.

Table 1.

Multiple regression analysis results for β -amyloid deposition, N = 54

Variable	B	SEB	β	p-value
Intercept	1.062	0.330		0.002*
Age	0.003	0.004	0.099	0.455
Sex (Male)	-0.065	0.049	-0.201	0.192
Statin use (+)	0.012	0.046	0.040	0.788
APOE4 carrier status (+)	0.085	0.043	0.256	0.054
HDL cholesterol	-0.001	0.001	-0.105	0.507
Non-HDL cholesterol	-0.001	0.001	-0.292	0.067
Resting CVCi	0.164	0.144	0.213	0.260
Resting CVCi x APOE4 carrier status (+)	-0.476	0.207	-0.433	0.026*

B = unstandardized regression coefficient; SE_B = standard error of the coefficient; β = standardized coefficient; *significant (p < 0.05), APOE4 = Apolipoprotein E4; HDL = High-Density Lipoprotein; CVCi = Cerebrovascular Conductance index

Figure 1. Relationship between β-amyloid deposition and resting Cerebrovascular Conductance index (CVCi) for APOE4 carriers and non-carriers (N = 54).

For APOE4 carriers, poor CVCi in the middle cerebral artery was associated with greater brain β-amyloid load. In contrast, β-amyloid load was not related to CVCi for APOE4 non-carriers. These findings suggest cerebrovascular dysfunction may promote AD-related brain pathology preferentially in APOE4 carriers.

To better understand the role of brain volumes on our selected variables of interest, we included gray and white matter volumes (mL) as independent variables in the multiple regression analysis. There was no significant effect of gray matter volume (p = 0.769) or white matter volume (p = 0.978) on β-amyloid deposition, and the interaction between resting CVCi and APOE4 carrier status remained significant (p = 0.032). This suggests the differential relationship between CVCi and β-amyloid by APOE4 carrier status is independent of differences in brain volume.

Cerebrovascular function during exercise

The TCD signal was lost for two participants during exercise leaving 52 participants in the exercise analysis (63% female, 31% APOE4 carriers; 70.8 ± 5.1 years old). We ran a multiple linear regression to predict ΔMCAv to moderate-intensity exercise from age, sex, β-amyloid deposition, statin use, APOE4 carrier status, HDL cholesterol, non-HDL cholesterol, the interaction between non-HDL cholesterol and APOE4 carrier status. There was linearity as assessed by partial regression plots and a plot of studentized residuals against the predicted values. There was independence of residuals, as assessed by a Durbin-Watson statistic near 2. There was no evidence of multicollinearity, as assessed by tolerance values greater than 0.1. There was homoscedasticity, as assessed by visual inspection of a plot of studentized residuals versus unstandardized predicted values. Data met the assumption of normality, as assessed by a P-P Plot. There were no outliers, as assessed by no standardized residuals greater than ±3 standard deviations (SD).

The multiple regression model significantly predicted ΔMCAv, F(8, 43) = 2.760, p = 0.015, R² = 0.339, adj. R² = 0.216. Regression coefficients and standard errors can be found in Table 2. There was a significant interaction (p = 0.014) between non-HDL cholesterol and APOE4 carrier status, suggesting the relationship between non-HDL cholesterol and ΔMCAv is moderated by the APOE4 allele. This relationship is shown graphically in Figure 2 for participants divided by APOE4 carrier status. For APOE4 carriers, there is a negative association between non-HDL cholesterol and ΔMCAv, with lower levels of non-HDL cholesterol predicting a more robust response to exercise in the cerebrovasculature. Conversely, the opposite relationship was apparent for APOE4 non-carriers, with higher non-HDL cholesterol predicting a larger ΔMCAv during moderate-intensity exercise. Male participants had a significantly larger ΔMCAv than female participants (p = 0.028). Contrary to our hypothesis, there was no significant effect of APOE4 carrier status on ΔMCAv (p = 0.454), suggesting the exercise response in the cerebrovasculature was not different between APOE4 carriers and non-carriers.

Table 2.

Multiple regression analysis results for the cerebrovascular response to moderate-intensity exercise (ΔMCAv), N = 52

Variable	B	SEB	β	p-value
Intercept	8.469	11.537		0.467
Age	-0.049	0.135	-0.047	0.720
Sex (Male)	3.916	1.728	0.359	0.028*
Beta-amyloid deposition	-6.592	4.790	-0.195	0.176
Statin use (+)	-0.665	1.518	-0.063	0.663
APOE4 carrier status (+)	1.183	1.567	0.104	0.454
HDL cholesterol	0.090	0.046	0.313	0.054
Non-HDL cholesterol	0.065	0.029	0.413	0.030*
Non-HDL cholesterol x APOE4 carrier status (+)	-0.118	0.046	-0.408	0.014*

B = unstandardized regression coefficient; SE_B = standard error of the coefficient; β = standardized coefficient;*significant (p < 0.05), APOE4 = Apolipoprotein E4; HDL = High-Density Lipoprotein

Figure 2. Relationship between the Cerebrovascular Response to exercise (ΔMCAv) and non-High-Density Lipoprotein (non-HDL) cholesterol levels for APOE4 carriers and non-carriers (N = 52).

For APOE4 carriers, higher non-HDL cholesterol levels were associated with a blunted ΔMCAv during an acute bout of moderate-intensity exercise. In contrast, higher non-HDL cholesterol predicted a larger ΔMCAv for APOE4 non-carriers. These findings suggest APOE4 moderates the association between pro-atherogenic cholesterol and cerebrovascular dysfunction.

There was no significant effect of gray matter volume (p = 0.658) or white matter volume (p = 0.691) on ΔMCAv with moderate-intensity exercise, and the interaction between non-HDL cholesterol and APOE4 carrier status remained significant (p = 0.015). This suggests the observed differential relationship between non-HDL cholesterol and ΔMCAv by APOE4 carrier status was independent of brain volume differences.

Discussion

In the current study, we found APOE4 carrier status moderated the association between vascular risk factors and brain pathology. Specifically, poor cerebrovascular conductance (CVCi, an indication of the brain’s perfusion capability)³² predicted higher brain β-amyloid load for the APOE4 carriers but not non-carriers. Additionally, although APOE4 carriers and non-carriers did not differ in overall cerebrovascular response to moderate-intensity exercise (ΔMCAv), higher non-HDL cholesterol (“bad” cholesterol)²⁵ was associated with a blunted response of the cerebrovasculature to moderate-intensity exercise for the APOE4 carriers, with the opposite relationship for non-carriers. Overall, these findings suggest vascular risk factors may differentially impact brain pathology for cognitively normal APOE4 carriers compared to non-carriers, providing insight into potential mechanisms of late-onset Alzheimer’s disease (AD) pathogenesis and reinforcing the importance of maintaining vascular health specifically for people at highest genetic risk of AD.

Cerebrovascular function at rest

A recent comprehensive study including over 7,700 brain scans from the Alzheimer’s Disease Neuroimaging Initiative database concluded cerebrovascular dysfunction was the strongest and earliest pathological abnormality in AD pathogenesis, followed next by β-amyloid deposition and later by brain metabolic dysfunction, functional MRI abnormalities and structural atrophy.¹ The relationship between cerebrovascular dysfunction and β-amyloid deposition is complex and likely bidirectional, with each contributing to the other at various stages of disease.¹⁵ However, there is growing evidence that cerebrovascular dysfunction can promote β-amyloid deposition. For example, experimental induction of brain hypoperfusion in mice via bilateral carotid artery stenosis significantly accelerates β-amyloid deposition.^33–35 The mechanisms through which hypoperfusion promotes β-amyloid deposition remain a matter of debate, but studies have suggested reduced blood flow may impair the dynamics of the interstitial fluid and result in congestion, which subsequently facilitates β-amyloid aggregation.³⁶ In addition to impairing clearance, low brain blood supply upregulates hypoxia-inducible factor, which increases β-secretase transcription leading to higher levels of β-amyloid production.^{37, 38} Furthermore, the addition of vascular risk factors to mouse models of AD has been shown to facilitate β-amyloid deposition by enhancing amyloidogenic amyloid precursor protein processing.³⁹ In humans, APOE4 carriers have been shown to have reduced CBF that precedes β-amyloid deposition, with lower CBF predicting increased β-amyloid accumulation, suggesting hypoperfusion may occur upstream of β-amyloid deposition.¹⁶

In the present study, we found lower conductance (CVCi) in the middle cerebral artery, the largest conduit vessel in the brain, predicted elevated β-amyloid load for the APOE4 carriers but not non-carriers. Since cerebrovascular conductance reflects the degree of cerebral blood flow (CBF) per perfusion pressure,³² this means APOE4 carriers who had lower resting CBF per a given perfusion pressure had significantly greater β-amyloid load, while this was not the case for non-carriers. Considering the updated pathogenesis timeline model of AD and the increasing evidence for a direct effect of cerebrovascular dysfunction on promoting β-amyloid deposition, these findings suggest cerebrovascular dysfunction may preferentially promote AD pathology in APOE4 carriers. This finding is in line with prior studies showing a synergistic effect between the APOE4 allele and cerebrovascular risk factors such as stroke.¹¹

Cerebrovascular function during exercise

The product of the APOE4 gene functions primarily in cholesterol transport throughout the body.⁴⁰ APOE4 significantly increases the risk of dyslipidemia, coronary artery disease, myocardial infarction, and ischemic/hemorrhagic stroke, in addition to AD.⁷ Moreover, some studies have suggested these cardiovascular diseases are not only more common in APOE4 carriers but may also act synergistically with the APOE4 allele when they occur.^{8–13, 40} Additionally, higher cholesterol has been associated with faster rates of cognitive decline in older adults with and without AD.^{5, 13} In the present study, we sought to investigate this relationship further, specifically analyzing the interaction between non-HDL cholesterol (colloquially known as “bad cholesterol”)²⁵ and the APOE4 allele on the ΔMCAv with exercise. Additionally, considering growing evidence that the APOE4 allele causes cerebrovascular dysfunction,^{3, 9, 16, 40} we hypothesized APOE4 carriers would have a blunted cerebrovascular response to exercise (ΔMCAv) compared to non-carriers. Importantly, the inclusion of ΔMCAv in the present study expands upon the resting CVCi findings by allowing for characterization of cerebrovascular function during dynamic physiological challenge (in this instance, aerobic exercise) to the brain vasculature, which may differ from the resting state. Previous studies reported dynamic cerebrovascular dysfunction in APOE4 carriers in response to physiological stimuli such as hypercapnia (cerebrovascular reactivity)⁴¹ and cognitive tasks.⁴² In the present study, we utilized ΔMCAv with a moderate-intensity exercise stimulus because this reflects the cerebrovascular response to a variety of stimuli and is relevant to daily life. That is, during aerobic exercise the cerebrovasculature is challenged by simultaneous alterations in arterial blood gas, perfusion pressure, metabolism and neuronal activity.¹⁸ Characterizing the response to this dynamic physiological challenge can therefore provide novel insight into cerebrovascular dysfunction that may not be observed at rest or with other physiological stimuli. We previously demonstrated a blunted ΔMCAv during exercise with increasing age,¹⁹ after ischemic stroke,^{20, 21} and in people with elevated β-amyloid.²² In the current manuscript, we sought to investigate whether the cerebrovascular response to exercise (ΔMCAv) varies as a function of APOE4 carrier status with the hope of providing further insight into pathogenic mechanisms of AD for those at highest genetic risk.

Contrary to our hypothesis, we found no difference in ΔMCAv from rest to exercise between APOE4 carriers and non-carriers, when controlling for age, sex, β-amyloid deposition and statin use. This may be due to the fact that our study included only participants with normal cognition and no history of diseases commonly associated with cardiovascular dysfunction, such as diabetes and heart failure. That is, the APOE4 carriers in our sample may represent a group that is unusually healthy compared to the general population, which could explain why these participants had normal cognition despite significantly elevated AD risk. However, while the overall sample may be too healthy to observe group differences between APOE4 carriers and non-carriers, the interaction term in the model allows us to investigate how certain risk factors (as they vary within the sample population) differentially relate to outcomes depending on participant genotype. To this end, we found a significant interaction between non-HDL cholesterol and APOE4 carrier status on ΔMCAv. This suggests the relationship between non-HDL cholesterol, a comprehensive measure of pro-atherogenic particles in the blood,^{14, 24–26, 31} and ΔMCAv is different between APOE4 carriers and non-carriers. For APOE4 carriers, higher non-HDL cholesterol level was associated with a blunted ΔMCAv from rest to exercise, which could argue for a role for this vascular risk factor in promoting cerebrovascular dysfunction specifically in APOE4 carriers. Intriguingly, the relationship between non-HDL cholesterol and ΔMCAv occurred in the opposite direction for non-carriers, with higher levels of non-HDL cholesterol predicting a more robust ΔMCAv with exercise. The mechanism behind this unexpected finding is unclear but may indicate a differential role for blood lipid components in cerebrovascular function for APOE4 carriers and non-carriers and merits further exploration.

Our findings remained significant when controlling for brain volume (both gray and white matter) in the multiple regression models, suggesting the observed relationships exist independently of brain volume. This result is consistent with most current models of AD pathogenesis, which place CBF decline and β-amyloid deposition upstream of structural MRI changes^{1, 2, 43} For example, one recent study using over 7,700 brain scans from the Alzheimer’s Disease Neuroimaging Initiative determined cerebrovascular dysregulation occurs earliest in AD pathogenesis, followed by β-amyloid deposition which is then followed by brain metabolic and functional abnormalities, all of which occur before structural atrophy is observed.¹ The fact that the relationship between poor CVCi and β-amyloid deposition occurred independently of brain volume differences in our sample provides further support for this model. Likewise, the interaction between non-HDL cholesterol and ΔMCAv remained significant when controlling for white and gray matter volume, and there was no relationship between these volumes and the ΔMCAv with moderate-intensity exercise. These results parallel current models of AD that posit cerebrovascular abnormalities occur earlier in the disease process than structural atrophy^{1, 2} and may further support the hypothesis that cerebrovascular dysregulation plays an early, causal role in AD pathogenesis.

Importantly, if vascular risk factors act synergistically with APOE4 to cause dementia, then interventions that improve cardiovascular health may be more effective for maintaining brain health and preventing cognitive decline in APOE4 carriers than non-carriers. Indeed, there is evidence that exercise - one of the most potent interventions to improve cardiovascular health - may preferentially benefit APOE4 carriers. For example, physical activity and exercise have been shown to more robustly prevent cognitive decline in APOE4 carriers than non-carriers,^{44, 45} and physical activity may preserve hippocampal volume selectively in APOE4 carriers.⁴⁶ Our current findings provide further evidence for a potential role for improving systemic cardiovascular health in order to improve brain function and potentially prevent or delay cognitive decline in APOE4 carriers.

Limitations

This study has important limitations. First, transcranial Doppler ultrasound (TCD) allows us to assess dynamic cerebrovascular function in ways that are not possible in an MRI scanner,⁴⁷ but TCD requires the assumption of constant artery diameter in order for velocity to serve as a surrogate for flow, which has limitations.^{48, 49} We included statin use in both regression models to control for potential impact of this lipid-lowering medication, but we treated this as a categorical variable (“yes” or “no” for statin use) which does not account for dosage or treatment duration. Cholesterol levels were obtained during a clinic visit 1-2 months before the vascular laboratory visit because we were interested in obtaining a rough measure of dyslipidemia (reflective of real-world clinical care) rather than assessing the acute effects of cholesterol on the cerebrovasculature. However, future studies could provide additional insight by including cholesterol measurements on the day of the vascular visit. Our participants were cognitively normal older adults which may limit generalizability to patient populations. Finally, the cross-sectional design of the study makes it impossible to definitively elucidate cause and effect (for example, whether low CVCi contributes to β-amyloid deposition or vice versa), and future studies are necessary to establish directionality.

Conclusion

In the current study, we found APOE4 carrier status moderates the association between central and peripheral vascular risk factors and brain pathology. APOE4 carriers with poor conductance in the cerebral vasculature demonstrated a greater brain β-amyloid load, while this relationship was not apparent for APOE4 non-carriers. Conversely, APOE4 carriers with better conductance in the largest conduit vessel supplying blood to the brain had significantly lower brain β-amyloid deposition. Although APOE4 carriers did not demonstrate an overall blunted response to exercise in the cerebrovasculature (lower ΔMCAv), APOE4 was found to moderate the association between non-HDL cholesterol and ΔMCAv. Specifically, higher non-HDL cholesterol was associated with a blunted ΔMCAv in APOE4 carriers but with a more robust ΔMCAv in non-carriers. These findings provide potential mechanistic insight into the pathogenesis of AD and reinforce the importance of maintaining both peripheral and central vascular health for people at the highest known genetic risk of late-onset AD.