Cerebral blood flow (CBF) alterations (Bangen et al., 2014) and amyloid-β (Aβ) accumulation (Rodrigue et al., 2012) have been independently linked to increased risk of developing dementia. It is well established that Aβ accumulation is an early event in the Alzheimer’s disease (AD) pathological process (Jack et al., 2010, 2013) and there is accumulating evidence of the role of early cerebral vascular dysfunction in AD (Iadecola, 2004; Zlokovic, 2011). This includes disruptions in neurovascular function, which is the normal regulation of CBF by arterioles and the capillary neurovascular unit (Girouard and Iadecola, 2006).

Arterial spin labeling (ASL) is a non-invasive magnetic resonance imaging (MRI) technique in which arterial water is magnetically labeled and used as an endogenous tracer to measure CBF (Detre and Alsop, 1999). ASL has been used to reliably measure CBF in AD patients (Johnson et al., 2005); individuals with mild cognitive impairment (MCI) (Bangen et al., 2012); and cognitively normal older adults (Bangen et al., 2009). ASL studies of individuals with AD demonstrate similar patterns of regional hypoperfusion as those shown with studies using fluorodeoxyglucose positron emission tomography (FDG-PET) and single photon emission computed tomography (SPECT) (Chen et al., 2011; Takahashi et al., 2014). ASL techniques have advantages over PET and SPECT including (1) non-invasive use of an endogenous tracer rather than an intravenously administered contrast agent; (2) relatively brief scan times (typically 5–10 min) and can be repeated in short succession due to the magnetization of the labeled blood water that decays within seconds; and (3) quantitative measurement of CBF at rest or during a functional task (Johnson et al., 2005). These advantages along with its increased sensitivity and ability to quantitatively measure perfusion make it ideal to extend its applications for research and in clinical settings (Telischak et al., 2015) designed to monitor neural and vascular changes in healthy aging and disease.

Previous studies have reported associations between Aβ deposition and CBF among older adults across the cognitive spectrum from normal aging to AD. For example, among 182 Alzheimer’s Disease Neuroimaging Initiative (ADNI) participants, Mattsson et al. (2014) reported that higher cortical Aβ load measured by florbetapir PET imaging was associated with reduced CBF in several regions of interest, independent of diagnostic group (cognitively normal, early MCI, late MCI, or AD) (Mattsson et al., 2014). Further, they reported that associations of Aβ load with CBF and brain volume varied across the disease stages. Specifically, in normally aging participants, higher Aβ load was associated with reduced CBF; however, in individuals with late MCI and dementia, higher Aβ load was related to greater reductions of gray matter volume. Given these findings, it was hypothesized that Aβ pathology may lead to reduced CBF early in the disease process and volumetric changes later in the disease process, although longitudinal studies are needed to confirm these temporal relationships (Mattsson et al., 2014). In another study including a sample of 27 cognitively normal older adults and 16 individuals diagnosed with amnestic MCI, Michels et al. (2016) reported a trend toward lower global CBF among those that had greater Aβ deposition measured with Pittsburgh Compound B (PiB) PET (Michels et al., 2016). Taken together, these studies suggest that CBF may be an important mechanism leading to cognitive decline, and may play an even more prominent role among those with elevated Aβ load.

Findings from several postmortem studies (Arriagada et al., 1992; Ingelsson et al., 2004) and in vivo PET imaging studies (Engler et al., 2006; Jack et al., 2009) have found no significant association between fibrillar amyloid load and degree of cognitive impairment in individuals with AD dementia. As such, it is thought that fibrillar aggregates of Aβ may not be the immediate cause of cognitive decline and/or Aβ accumulation may be an early event in the AD pathological cascade and may plateau before onset of dementia (Hampel, 2013). If Aβ accumulation is most dynamic before onset of dementia, its effects on cognition should be studied prior to the onset of significant cognitive decline (Hampel, 2013). Although several postmortem and amyloid PET studies have shown that a considerable portion of asymptomatic older adults have increased Aβ burden in the absence of any cognitive impairment (Price and Morris, 1999; Fagan et al., 2006), other previously published reports have shown statistically significant associations between increased Aβ load on PET and poorer cognitive performance in cognitively normal older adults (Rentz et al., 2011). These effects may be best detected on challenging episodic memory tasks and may interact with various AD risk factors such as genetic risk (Pike et al., 2011; Kantarci et al., 2012). Little is known about how CBF and Aβ, which may both serve as early markers of AD changes, may interact to affect cognition in cognitively normal older adults.

There is growing evidence supporting the notion that ASL MRI may be a useful biomarker in predicting cognitive decline and progression to MCI and dementia (Chao et al., 2010; Beason-Held et al., 2013). However, most previous studies have focused on individuals already demonstrating cognitive impairment (MCI and AD) and, to our knowledge, no study has considered how ASL MRI CBF and Aβ status may interact to affect cognition in cognitively normal older adults. Therefore, we examined potential statistical interactions of ASL MRI CBF and Aβ status on cognitive function within a sample of normally aging older adults drawn from the ADNI study.

Our study extends previous CBF studies of dementia risk by showing statistical interactions between Aβ status (negative or positive) and regional CBF on memory performance in a sample of well-characterized, cognitively normal older adults. Specifically, we found that among Aβ positive older adults, there were significant associations between higher CBF and poorer verbal memory performance in regions known to be predilections sites for AD—the hippocampus, posterior cingulate, and precuneus. In contrast, among Aβ negative older adults, there were no significant relationships between memory performance and CBF, although there was a trend toward higher CBF in the posterior cingulate and better verbal memory performance. Importantly, our findings demonstrate differential associations between CBF and cognition for Aβ positive versus negative cognitively normal older adults.

Although regional decreases in CBF are interpreted as reflecting decreased brain function, increases in perfusion in the context of preclinical AD—particularly when cognitive performance is maintained or even improved—has often been considered to represent a compensatory response to an incipient pathologic process (Dai et al., 2009). Indeed, several previously published studies have found significant differences in resting hyperperfusion in tandem with better memory function in non-demented older adults at risk for AD, and researchers have interpreted this finding as a potential compensatory response reflecting metabolic alterations and/or increased need for glucose and oxygen to support neuronal activity (Fleisher et al., 2009; Bangen et al., 2012; Zlatar et al., 2014). In contrast, we found that higher resting CBF was associated with poorer memory performance among older adults at increased risk for AD by virtue of elevated Aβ accumulation, possibly reflecting cerebrovascular dysregulation or a cellular and/or vascular compensatory response to pathologic processes whereby higher CBF is needed to maintain normal memory abilities. Unlike our previously published work, all individuals in this study were cognitively normal and, importantly, there were no group differences among the Aβ positive and Aβ negative group in terms of cognitive performance. The heightened CBF in Aβ positive individuals may suggest that these individuals are on a declining trajectory of RAVLT performance (albeit still normal), and they need more CBF to support this declining memory system. Hyperperfusion in early MCI followed by hypoperfusion later in MCI when approaching the transition to dementia has been shown and it is possible that the Aβ positive individuals in our sample are closer to developing MCI. Further longitudinal studies investigating perfusion differences across the course of the disease are needed to further examine the role of higher CBF.

Our work showing statistical interactions of perfusion and Aβ status is consistent with previous studies that have demonstrated links between Aβ and cerebrovascular dysregulation. Specifically, prior work has shown that Aβ increases the vulnerability of the brain to cerebral ischemia through its effects on the cells of the neurovascular unit (Zhang et al., 1997; Iadecola, 2004; Girouard and Iadecola, 2006). Moreover, cerebrovascular dysfunction upregulates amyloid precursor protein and Aβ cleavage (Abe et al., 1991; Yokota et al., 1996; Iadecola, 2004). Ultimately, Aβ and cerebrovascular dysfunction are thought to reinforce one another thereby amplifying their deleterious effects on the brain (Iadecola, 2004). The present findings provide further support for the role of vascular alterations in the AD prodrome.

Previous studies of cerebral perfusion across the continuum from the preclinical phase to AD suggest a biphasic pattern characterized by early hyperperfusion preceding later hypoperfusion (Wierenga et al., 2014). In this way, cerebrovascular dysregulation becomes more pronounced over time as the disease progresses (Mentis et al., 1998). This may be due to several factors including neuronal death and synaptic loss resulting in a reduced hemodynamic response to neural activation; accumulating amyloid in cerebral arterioles leading to disruptions in the ability of vascular smooth muscles cells to relax thereby creating a mechanical obstacle to vasodilation (Christie et al., 2001); and atherosclerosis in the circle of Willis (Roher et al., 2003) and conduit cerebral arteries resulting in reduced global CBF and further disruption in the ability of neural stimuli to increase perfusion (Iadecola, 2004). Furthermore, evidence suggests that increased activation within neural networks may modulate Aβ accumulation given that brain regions with lifelong high activity levels (e.g., default mode network) also have the greatest predisposition for Aβ accumulation and increased synaptic transmission results in increased interstitial fluid Aβ levels (Cirrito et al., 2008; Hampel, 2013).

Accumulating evidence suggests that ASL CBF represents a useful biomarker in at-risk individuals since this technique can sensitively differentiate those at risk from control participants (Fleisher et al., 2009; Bangen et al., 2012; Wierenga et al., 2012). Additionally, ASL CBF indices have reliably predicted progression from normal cognition to MCI (Beason-Held et al., 2013), and MCI to AD (Chao et al., 2010). Longitudinal studies have shown that, relative to individuals who remained cognitively normal, older adults who later developed MCI demonstrated hyperperfusion in orbitofrontal, medial frontal, and anterior cingulate regions over time, accompanied by reduced CBF in parietal, temporal, and thalamic regions (Beason-Held et al., 2013). These changes occurred several years prior to the development of cognitive impairment and were observed in regions known to be predilection sites for early AD pathology (Beason-Held et al., 2013). Additionally, these changes were independent of longitudinal changes in tissue volume. This is consistent with findings from our secondary analyses that revealed significant interactions of Aβ status and CBF on memory performance independent of volume or cortical thickness, further suggesting that CBF may play a role in cognitive functioning independent of tissue loss.

In the few existing longitudinal prospective studies using ASL MRI, resting hypoperfusion of the right inferior parietal cortex and right middle frontal cortex at baseline predicted progression from MCI to dementia at 3-year follow-up (Chao et al., 2010) and in another study reduced CBF in the posterior cingulate at baseline was associated with development of cognitive decline at 18-month follow up in healthy older adults (Xekardaki et al., 2015). Our present findings highlight the important association between CBF and memory, and they provide further support for the notion that CBF is a useful marker of AD risk and correlate of cognitive function in older adults. Specifically, we observed evidence of dysregulated CBF patterns in Aβ positive individuals who are cognitively normal suggesting that ASL MRI is sensitive to very early changes in the brain.

The present findings suggest that Aβ accumulation and CBF alterations together influence memory performance in at-risk older adults. These findings add to a growing body of evidence underscoring the importance of multiple pathological processes co-occurring in AD and the interactive influence of several risk factors. Neuropathological studies have shown that clinically diagnosed MCI and AD are both pathologically heterogeneous disorders (Schneider et al., 2007; Nettiksimmons et al., 2014). In our own sample of autopsy-confirmed AD, we found that the presence of mild cerebrovascular changes was associated with less severe AD pathology yet there were no differences in severity of cognitive impairment between the AD patients with and without evidence of cerebrovascular disease (Bangen et al., 2015). These results raise the possibility that cerebrovascular changes contribute to overall severity of cognitive impairment, even in patients with both autopsy-confirmed AD and relatively mild cerebrovascular disease (Bangen et al., 2015). We have also shown that the presence of multiple AD risk factors (e.g., advanced age, APOE 𝜀4 allele, family history of AD, and/or increased vascular risk burden in different combinations) has additive or interactive effects on brain function and cognition (Fleisher et al., 2009; Bangen et al., 2014). The present findings extend this work by demonstrating interactions between PET brain Aβ positivity and CBF on memory performance.

Our results did not reveal any significant main effects of Aβ status or CBF on memory performance in our sample. With respect to Aβ, findings from cross-sectional studies have been inconsistent with some studies reporting no relationship between burden of amyloid in the brain and cognition in cognitively normal or non-demented older adults (Mintun et al., 2006; Aizenstein et al., 2008; Mormino et al., 2009; Rowe et al., 2010) whereas other studies showed associations between greater amyloid and worse cognition (Rodrigue et al., 2012). Additionally, other studies have showed relationships between greater amyloid and worse cognition in APOE 𝜀4 carriers, while no such relationship (Lim et al., 2013) or a weaker relationship among non-carriers (Kantarci et al., 2012).

Prospective longitudinal studies have also been mixed with some studies reporting greater faster rates of cognitive decline in non-demented older adults with high cerebral Aβ load over an 18-month period following PET imaging (Doraiswamy et al., 2012; Lim et al., 2012; Ellis et al., 2013; Kawas et al., 2013) whereas other studies have found no difference in rate of cognitive change over 2- to 3-year follow-up between cognitively normal older adults who had high versus low Aβ at baseline (Villemagne et al., 2011; Ewers et al., 2012). However, prospective longitudinal studies have generally had little follow up after PET imaging (Gu et al., 2015), and retrospective longitudinal studies have shown that non-demented older adults who have higher levels of Aβ showed faster cognitive decline prior to PET scanning relative to their counterparts with lower level of Aβ (Resnick et al., 2010; Landau et al., 2012; Gu et al., 2015). A meta-analysis of 64 studies examining amyloid-cognition associations in healthy older adults found that episodic memory had a small and significant relationship to amyloid burden whereas other cognitive abilities (e.g., working memory, processing speed, visuospatial function, semantic memory) did not have significant relationships to amyloid. Study design, that is cross-sectional vs. longitudinal design, had little influence on findings (Hedden et al., 2013). Although the role of Aβ in cognitive decline and the clinical expression of AD is complex and may be moderated by additional risk factors and variables (Kantarci et al., 2012; Lim et al., 2013; Gu et al., 2015), there is clear evidence to suggest it contributes to the AD process and pivotal to the amyloid cascade model (Jack et al., 2010, 2013).

In contrast to our current ADNI-based findings, we have previously found in our own community samples main effects of CBF on cognition when examining both cognitively normal older adults and those with MCI (Bangen et al., 2012, 2014). In the present sample of ADNI participants, all individuals were cognitively normal and, given selection criteria for ADNI, all participants had very low vascular risk burden. It is possible that we would have found main effects of CBF on memory if there were a greater range of cognitive performance and CBF values. A previously published paper in the ADNI cohort found that the effects of higher brain Aβ load was associated with reduced CBF in cognitively normal older adults and with reduced brain volume in late MCI and dementia suggesting that the relationship between Aβ and CBF changes over the course of the disease (Mattsson et al., 2014). In the current study, we focused on the interaction between Aβ and CBF on memory and it is possible that we would have observed different relationships among Aβ status, CBF, and memory performance if we included participants with more pronounced cerebrovascular disease and/or individuals with MCI or AD. However, given a critical need to examine preclinical AD in its very earliest stages, for the purposes of the current study we emphasized associations among Aβ status, CBF and memory function in older adults who show brain Aβ positivity on PET in the context of no detectable cognitive impairment.

This work has several important research and clinical implications. First, our findings suggest a dynamic relationship between cerebral perfusion and Aβ in the expression of memory function in individuals with preclinical AD. Results further underscore the potential value in examining sensitive vascular variables in the pathogenesis of AD. Additionally, pharmacological and behavioral interventions, including physical exercise, may play a critical role in the regulation of CBF and, ultimately, the prevention of cognitive decline. Interestingly, a recent study showed that older adults taking angiotensin II AT1-receptor blockers exhibited reduced cerebral amyloid retention (Nation et al., 2016). As noted by the authors, this finding is consistent with results from studies in transgenic animals, and they may explain in part why older adults who use AT1-receptor blockers show reduced progression to dementia despite greater vascular risk burden (Nation et al., 2016). Future research is needed to further determine whether anti-hypertension medication and/or behavioral lifestyle changes may improve cerebral microcirculation and reduce Aβ retention.

Strengths of this study include a well-characterized sample of older adults who have undergone multi-modal neuroimaging and neuropsychological assessment as part of a national study on aging and AD. Limitations of our study include use of a global measurement of Aβ pathology rather than local or regional measures. In addition, this was a cross-sectional study and we did not assess cognitive outcome. It is possible that some of the Aβ positive individuals in this study will not develop AD and, likewise, some of the Aβ negative individuals may express the disease at some point. Furthermore, previously published results have reported an absence of cross-sectional associations between amyloid and cognition in healthy controls but have found negative associations for when data is examined longitudinally (Gu et al., 2015). Despite these limitations, in the search for reliable biomarkers of very early AD, ASL MRI may prove especially useful, and the combination of both cerebrovascular and Aβ markers may more completely inform the complex pathological processes underlying the clinical expression of AD than either biomarker class alone. Finally, since vascular risk factors are modifiable, these results may have important implications for biomarker studies, clinical trials, and treatment.

KB designed the study, analyzed and interpreted the data, and wrote and revised the manuscript. AC analyzed and interpreted the data and wrote and revised the manuscript. EE, NE, MW, KT, LL, MT, ZZ, DN, MB, and LD-W interpreted the data and revised the manuscript for important intellectual contact. All authors approved the submitted version of the manuscript and agree to be accountable for all aspects of the work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.