Cerebrovascular disease promotes tau pathology in Alzheimer’s disease

Abstract

Small vessel cerebrovascular disease, visualized as white matter hyperintensities on T2-weighted magnetic resonance imaging, contributes to the clinical presentation of Alzheimer’s disease. However, the extent to which cerebrovascular disease represents an independent pathognomonic feature of Alzheimer's disease or directly promotes Alzheimer’s pathology is unclear. The purpose of this study was to examine the association between white matter hyperintensities and plasma levels of tau and to determine if white matter hyperintensities and tau levels interact to predict Alzheimer’s disease diagnosis. To confirm that cerebrovascular disease promotes tau pathology, we examined tau fluid biomarker concentrations and pathology in a mouse model of ischaemic injury. Three hundred ninety-one participants from the Alzheimer’s Disease Neuroimaging Initiative (74.5 ± 7.1 years of age) were included in this cross-sectional analysis. Participants had measurements of plasma total-tau, cerebrospinal fluid beta-amyloid, and white matter hyperintensities, and were diagnosed clinically as Alzheimer’s disease (n = 97), mild cognitive impairment (n = 186) or cognitively normal control (n = 108). We tested the relationship between plasma tau concentration and white matter hyperintensity volume across diagnostic groups. We also examined the extent to which white matter hyperintensity volume, plasma tau, amyloid positivity status and the interaction between white matter hyperintensities and plasma tau correctly classifies diagnostic category. Increased white matter hyperintensity volume was associated with higher plasma tau concentration, particularly among those diagnosed clinically with Alzheimer’s disease. Presence of brain amyloid and the interaction between plasma tau and white matter hyperintensity volume distinguished Alzheimer’s disease and mild cognitive impairment participants from controls with 77.6% and 63.3% accuracy, respectively. In 63 Alzheimer’s Disease Neuroimaging Initiative participants who came to autopsy (82.33 ± 7.18 age at death), we found that higher degrees of arteriosclerosis were associated with higher Braak staging, indicating a positive relationship between cerebrovascular disease and neurofibrillary pathology. In a transient middle cerebral artery occlusion mouse model, aged mice that received transient middle cerebral artery occlusion, but not sham surgery, had increased plasma and cerebrospinal fluid tau concentrations, induced myelin loss, and hyperphosphorylated tau pathology in the ipsilateral hippocampus and cerebral hemisphere. These findings demonstrate a relationship between cerebrovascular disease, operationalized as white matter hyperintensities, and tau levels, indexed in the plasma, suggesting that hypoperfusive injury promotes tau pathology. This potential causal association is supported by the demonstration that transient cerebral artery occlusion induces white matter damage, increases biofluidic markers of tau, and promotes cerebral tau hyperphosphorylation in older-adult mice.

Plasma-derived total-tau concentration is highly correlated with WMH volume, an MRI marker of small vessel ischaemic damage, particularly among individuals with clinical Alzheimer’s disease. A parallel mouse model confirmed white matter damage, hyperphosphorylated tau pathology and elevated tau concentration in plasma and CSF due to surgically induced ischaemic injury.

Graphical Abstract

Graphical Abstract.

Introduction

Models of Alzheimer’s disease pathogenesis emphasize prescribed pathological changes that begin with beta-amyloid deposition, tau pathology and associated neurodegeneration (Jack et al., 2013, 2018). However, in recent years, there has been an increased appreciation of the contribution of vascular and cerebrovascular disease (CVD) to the clinical expression of Alzheimer’s disease. Epidemiologic data emphasize vascular factors, such as hypertension, mid-life obesity and diabetes in the increased risk for clinical Alzheimer’s disease (Barnes and Yaffe, 2011). Neuroimaging and pathological studies highlight that CVD co-occurs with primary Alzheimer’s disease pathology and drives disease expression more often than not (Schneider et al., 2004; Kapasi et al., 2017). There is also a clear ‘shared genetic risk’ for CVD and Alzheimer’s disease with several cardiovascular-associated genes increasing the risk for clinical Alzheimer’s disease (Broce et al., 2019). These studies have generally been interpreted as evidence that CVD is a common comorbidity that contributes additively to the clinical symptoms of Alzheimer’s disease. However, the question of whether small vessel CVD promotes Alzheimer’s disease pathology directly remains critical. A primary, pathogenic role of CVD in Alzheimer’s disease-related pathology and neurodegeneration could help to identify novel disease-modifying therapeutic targets.

Small vessel CVD manifests on magnetic resonance imaging (MRI) T2-weighted sequences as white matter hyperintensities (WMH) (Wardlaw et al., 2015). Joining other studies that implicate WMH in Alzheimer’s disease (Yoshita et al., 2006; Mimenza-Alvarado et al., 2018), we showed previously that higher WMH volume is associated with increased risk of incident clinical Alzheimer’s disease (Brickman et al., 2012, 2015), rate of cognitive decline among individuals with Alzheimer’s disease (Brickman et al., 2008), and the APOE-ɛ4 allele (Brickman et al., 2014). Individuals with autosomal dominant, fully penetrant forms of Alzheimer’s disease manifest increased WMH above non-mutation carrier control participants up to 20 years prior to the expected age of clinical onset, establishing WMH as a ‘core feature’ of Alzheimer’s disease (Lee et al., 2016).

Whether WMH—or the underlying small vessel ischaemic CVD they represent—cause accumulating Alzheimer’s disease pathology remains an open question. Most studies show no reliable relationship between WMH or other markers of white matter abnormality and amyloid pathology (Hedden et al., 2012) (cf. Grimmer et al., 2012), but emerging animal studies suggest that cerebral hypoperfusion or ischaemia do indeed increase tau hyperphosphorylation (p-tau) (Raz et al., 2019), which may be mediated by local inflammatory changes (Jalal et al., 2015). It is difficult to test these types of Alzheimer’s disease marker associations in humans because cerebrospinal and positron emission tomography biomarkers are invasive, costly, and difficult to acquire in adequate numbers of participants with MRI data to test hypotheses with sufficient statistical power. However, ultrasensitive enzyme-linked immunosorbent assay techniques have been developed to assay tau levels in the plasma (Rissin et al., 2010a). While generally considered inadequate for diagnostic purposes, there is consensus that plasma tau levels reflect Alzheimer’s disease-associated cerebral tau pathology because they are elevated among participants with clinical Alzheimer’s disease compared with unimpaired controls; correlate positively with cerebrospinal tau levels and negatively with amyloid levels; and predict cognition difficulties, cognitive decline, brain atrophy and decreasing cortical glucose metabolism (Mattsson et al., 2016; Pase et al., 2019). There are other compelling reasons to consider plasma measures when examining the association between cerebrovascular and Alzheimer’s disease makers. Plasma tau measures relate specifically to markers of small vessel CVD examined at autopsy, including microinfarcts (Pase et al., 2019). Compared with central markers of tau pathology, plasma total tau measures may also most accurately reflect damage related to ischaemia induced blood–brain barrier breakdown, which would allow smaller fragments of tau protein to pass into the blood than what remains in the parenchyma or cerebrospinal fluid (CSF) (Fossati et al., 2019).

Here, we tested the hypothesis that increased WMH volume is associated with increased tau pathology, indexed by plasma levels, in the Alzheimer’s Disease Neuroimaging Initiative (ADNI). A second hypothesis was that the strength of the relationship between WMH and tau levels increases as a function of clinical diagnostic category [i.e. normal control (NC), mild cognitive impairment (MCI) and clinical Alzheimer’s disease], reflecting the possibility that small vessel disease drives tau pathology, which has a subsequent impact on cognition and function. We further examined the extent to which WMH severity and its interaction with tau pathology predict clinical diagnosis. Additionally, in order to determine whether the observations of an association between WMH and plasma tau measures suggests a relationship between small vessel CVD and tau pathology, we examined the correlation between the severity of arteriosclerosis and neurofibrillary tangle pathology in a group of participants who came to autopsy.

To confirm a causal effect of CVD on tau pathology, we used a well-established rodent model of transient occlusion with reperfusion (Barone et al., 2009, 2012) and determined if brain ischaemia–reperfusion causes white matter damage, increased plasma and CSF tau concentrations, and cerebral tau hyperphosphorylation. This model produces myelin damage that mediates aspects of cognitive decline in animal models (Zhou et al., 2013; Liu et al., 2018). The aetiology of WMH is likely heterogeneous, and factors that have been implicated consistently include ischaemia, hypoxic damage and/or hypoperfusion (Wardlaw et al., 2015). We thus consider this model, particularly when implemented in older mice, as a relevant approach that can be used to test the hypothesis that small vessel CVD induces tau pathology. When combined with the human data, the present mouse studies provide initial experimental evidence that brain ischaemia, reflected in humans as WMH, induces tau pathology that is independent of beta-amyloid.

Materials and methods

Alzheimer’s Disease Neuroimaging Initiative

Human data used in the preparation of this article were obtained through the ADNI database (www.adni.loni.usc.edu), which was launched in 2003 as a public–private partnership and is led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial MRI, positron emission tomography, other biological markers, and clinical and neuropsychological assessments can be analysed and related to measures of the progression from NC to MCI and early Alzheimer’s disease dementia. All participating institutions received appropriate ethical approval from corresponding institutional review boards and ethics committees. Written informed consent was obtained from all study participants involved or their legal surrogates. Up-to-date information is available at www.adni-info.org.

Participants

We used ADNI data accessed on 22 June 2018. The study data and samples were collected from 7 September 2005 to 18 September 2007. Data were obtained for all participants classified as cognitively healthy controls, having MCI, or having Alzheimer’s disease dementia. Inclusion in the current study was limited to participants who also had available baseline derived plasma levels of tau and CSF measures of Aβ₄₂ (n = 391). General ADNI exclusion criteria included significant vascular disease risk history, defined as a Modified Hachinski Ischemia scale greater than 4. Recruitment and diagnostic procedures were described previously (Petersen et al., 2010). Characteristics and demographics for participants in this study are included in Table 1. Neuropathological data were obtained for 63 ADNI participants, 27 of whom were included in the antemortem analyses.

Table 1.

Demographic characteristics, amyloid status and plasma tau levels for participants included in the study

	NC	MCI	AD	Total	Statistic
N	108	186	97	391	–
Age, mean (SD) years	75.17 (5.28)	73.88 (7.51)	74.75 (7.84)	74.45 (7.06)	F = 1.25, P = 0.288
Sex, n (%) women	54 (50.0%)	60 (32.3%)	42 (43.3%)	156 (39.3%)	χ2 = 9.59, P < 0.008
Amyloid positive, n (%)	40 (37%)	136 (73%)	90 (92%)	266 (68%)	χ2 = 77.2, P < 0.001
Plasma tau, mean (SD) pg	2.55 (1.33)	2.80 (1.72)	3.17 (1.36)	2.82 (1.55)	F = 4.17, P = 0.016

AD = clinical Alzheimer’s disease; MCI = Mild Cognitive Impairment; NC = Cognitively Normal Controls.

Neuroimaging acquisition

A standardized MRI protocol was validated and implemented across ADNI sites (Jack et al., 2008a). High-resolution T1-weighted volumetric magnetization prepared rapid gradient echo sequences were obtained in the sagittal orientation on 1.5 T MRI systems. An axial proton density/T2-weighted dual contrast fast spin-echo sequence was also acquired for detection of WMH (Jack et al., 2008a). All sites were required to pass a rigorous scanner validation and each subject scan acquisition included fluid-filled phantoms. Details of scan validations can be found on the ADNI site (www.adni.loni.usc.edu).

WMH quantification/volumetric data

Detailed WMH quantification methods have been provided previously (Schwarz et al., 2009). Briefly, T1-, T2- and proton density-weighted MRI scans were co-registered, brain extracted and corrected for bias field inhomogeneity (Decarli et al., 1996; Wolz et al., 2011). WMH were detected at each voxel based on the corresponding T1, T2 and proton density intensities, the prior probability of WMH, and the conditional probability of WMH based on the presence of WMH in neighbouring voxels. Total WMH volumes were derived by summing voxels identified as hyperintense (Schwarz et al., 2009). WMH volumes derived at the ADNI baseline visit were included in this study.

In addition to WMH volume, cerebral volumetric data were downloaded from ADNI. Volumetric data were derived with FreeSurfer (http://surfer.nmr.mgh.harvard.edu/; Dale and Sereno, 1993; Dale et al., 1999; Fischl et al., 1999a, b, 2001, 2002, 2004; Han et al., 2006; Jovicich et al., 2006; Ségonne et al., 2007) applied to T1-weighted volumetric MRI. We calculated relative total volume (total brain volume/total intracranial volume) and relative white matter volume (total white matter volume/intracranial volume) as indicators of total brain and total white matter atrophy, respectively.

Human CSF biomarkers

Cerebral spinal fluid samples were collected at study baseline using the microbead-based multiplex immunoassay INNO-BIA AlzBio3 research use only test (Fujirebio, Ghent, Belgium) and run on the Luminex platform (Shaw et al., 2009). Detailed methods of the validation study, assay technology and quality measurements are located on the ADNI site (www.adni.loni.usc.edu). Due to the variation in reagent lot kit measured concentrations, typical in research use only immunoassay tests, multiple batch runs were acquired per participant. The median values provided by ADNI, which reflect the average of values obtained from the multiple reagent lot kits, were used for the analyses reported here. The CSF samples were drawn at baseline assessment, but the analytical run was conducted on 9 March 2016. CSF measures of Aβ₄₂ were used to determine amyloid positivity, which was derived from the cut off value of 192 pg/ml (Shaw et al., 2011). Cerebral spinal fluid measures of total tau and phosphorylated tau were also obtained and used as secondary outcomes.

Human plasma tau measurements

Plasma tau measurements were obtained at baseline and analysed with the Quanterix ‘Single molecule Array’ (Simoa) HD1 analyzer and Human Total Tau kit, using a monoclonal antibody capture that responded to a linear epitope in all tau isoforms. Detailed methods and analyses have been described in previous publications (Mattsson et al., 2016, Rissin et al., 2010b; Randall et al., 2013). Further information on quantification can be found on the ADNI site (www.adni.loni.usc.edu).

Neuropathology validation

To examine whether observed relationships between WMH and plasma tau suggest a relationship between small vessel disease and tau pathology, we turned to neuropathological data collected in ADNI. The ADNI neuropathology protocol follows the National Institute on Aging-Alzheimer’s Association (NIA-AA) guidelines for the assessment of Alzheimer’s disease (Montine et al., 2012). Using standardized neuropathological outcomes derived from the NIA-AA protocol, an ‘ABC’ score for neuropathological change was generated by ADNI to include histopathologic assessments of amyloid β deposits using Thal phases (A), Braak staging of neurofibrillary degeneration (B) and Consortium to Establish a Registry for Alzheimer's Disease scoring of neuritic plaques (C), in addition to detailed methods of assessing co-morbid conditions such as vascular brain injury (Montine et al., 2012). We calculated a summary arteriosclerosis index, comprising the sum of scalar arteriosclerosis severity ratings (absent/none = ‘0’, mild = ‘1’, moderate = ‘2’, severe = ‘3’) across 23 brain regions, including middle frontal gyrus, anterior cingulate gyrus, precentral gyrus, superior and middle temporal gyri, inferior parietal lobe, occipital lobe, amygdala, entorhinal cortex, CA1, dentate gyrus, parahippocampal gyrus, atherosclerosis within the circle of Willis and arteriosclerosis in various subcortical regions. We examined the relationship between the summary arteriosclerosis index and neurofibrillary pathology, operationally defined as overall Braak staging. Further information on postmortem quantification procedures and brain regions sampled can be found on the ADNI site (www.adni.loni.usc.edu/methods/neuropath-methods/).

Mouse ischaemia–reperfusion model

Mouse use was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health [National Research Council (US) Committee for the Update of the Guide for the Care and Use of Laboratory Animals, 2011]. Reporting the data on these mice follows the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines for animal research (Kilkenny et al., 2011). The experimental and surgical protocol for this research was approved by the SUNY Downstate Institutional Animal Care and Use Committee under Protocol Number 15-10475. All surgery were done under deep isoflurane anaesthesia. Euthanasia was performed under deep isoflurane anaesthesia using exsanguination and whole-body perfusion. All efforts were made to minimize or eliminate animal suffering.

In order to confirm that cerebral hypoperfusion, a source of WMH, can induce tau pathology, we used a well-established 30- and 60-min transient middle cerebral artery occlusion (tMCAo) brain ischaemia–reperfusion model in mice. We performed tMCAo in older-adult mice (13.72 ± 3.67-months-old) in order to phenocopy the human condition more closely. The tMCAo procedure or sham surgery (Barone et al., 2012) was conducted on 13 female C57Bl/6 J (Jackson Labs; Bar Harbor, ME, USA) mice, grouping chosen at random, with eight receiving tMCAo (30 min, n = 4; 60 min, n = 4) and five receiving Sham surgery. Briefly, under 2% isoflurane anaesthesia, a monofilament suture (6.0, from Doccol Corp, Sharon, MA, USA) was inserted through the proximal external carotid artery, advanced into the internal carotid artery and positioned to occlude the origin of MCA (about 10 mm from the common carotid artery bifurcation). Mice recovered from anaesthesia, and after 30 or 60 min of occlusion they were re-anaesthetized with isoflurane and the intraluminal filament was withdrawn to allow reperfusion. Most tMCAo protocols in mice and rats use occlusion times of 60 min or 120 min followed by reperfusion to produce relatively consistent brain infarctions that affect about 30% of the cerebral hemisphere. Infarcted brain areas include somatosensory, motor and parietal cortices and the striatum (Zhou et al., 2011, 2013; Barone et al., 2012). Here, we used 30 min and 60 min tMCAo to evaluate a range of hypoperfusion injury and to better approximate the variable ischaemic damage seen in human WMH. Sham surgery animals were subjected to the same procedure, but without occlusion–reperfusion of the MCA. Body temperature was continuously maintained at 37.0 ± 0.5°C using the Heat Therapy Pump-1500 and Temperature Therapy Pad (Adroit Medical Systems, Louden, TN, USA) until animals recovered completely from anaesthesia. Animals were then placed in their home cages, monitored closely over the next 4 h and then daily for the rest of the study. Bodyweight was monitored throughout the experiments and changed minimally (<5%) due to surgery.

Perfusion and fixation

The animals were perfused as previously described (Hernández et al., 2014). Briefly, mice were perfused with phosphate-buffered saline followed by 4% paraformaldehyde/phosphate-buffered saline. The brains were harvested and post-fixed in PFA/phosphate-buffered saline and transferred to 30% sucrose in phosphate-buffered saline until the brains were isopycnic with the solution. The brains were then quickly frozen in n-methylbutane, cooled over dry ice and kept at −80°C for cryostat sectioning. Free-floating coronal brain sections (30 µm) were obtained with coordinates between −1.34 and −1.46 posterior to Bregma and immunostained as described below.

Immunohistochemistry

We harvested the mouse brains 12–15 days after surgery for immunohistochemical and immunofluorescence (IF) analysis of myelin basic protein (MBP), total-tau, and p-tau.

Light microscopy

These methods have been described previously (Hernández et al., 2014; Gutiérrez-Vargas et al., 2017). In short, after quenching, sections were washed with phosphate-buffered saline and incubated with the phospho-tau (Ser 202, Thr 205; {1:500}) monoclonal antibody (AT8; Invitrogen, Waltham, USA), anti-tau antibody (Tau5 {1:250} Abcam, Cambridge, USA,). Tau5 is a pan-tau antibody that labels monomeric and aggregated tau equally well (Carmel et al., 1996). AT8 recognizes phosphorylated tau in mouse and human (Biernat et al., 1992). Tauopathy in Alzheimer’s disease is usually staged with AT8 antibodies (Braak et al., 2006). Sections were then incubated with biotinylated secondary antibody (1:2000) washed and then incubated with Vectastain (Vector Laboratories, Burlington, Canada), followed by 3,3′-diaminobenzidine (ImmPACT, Vector Laboratories), which was used to stain sections for light microscopy examination. Immunostaining intensity was evaluated with image J (NIH, Bethesda, MD, USA). The images were converted to a binary system, and integrated densities (relative units, RU) were obtained for each image. The background was automatically subtracted from each image to quantify the relative intensity of the immunostaining. For details, see Nguyen and Nguyen (2013) and Gutiérrez-Vargas et al. (2017).

IF microscopy

For IF AT8 (1:500) and MBP (1:2000 polyclonal antibody, Invitrogen) were incubated overnight, followed by the secondary antibodies that included Alexa Fluor^® 488 conjugated Goat anti-mouse (1:1000) and Alexa Fluor 647-conjugated Goat anti-chicken (1:200; both from Abcam). Sections were mounted on glass slides with 4′,6-diamidino-2-phenylindole (DAPI) solution (Vectashield, from Vector laboratories). The omission of the primary antibodies resulted in no staining. The sections were photographed at 40× magnification and the images were used to evaluate the fluorescence intensity of MBP immunostaining in regions of interest located in the corpus callosum and stratum alveus. Images were obtained using the laser Olympus FV 1000 D microscope and IF was measure with Fluoview FV1000 software. P-tau was evaluated by calculating the total number cells immuno-stained with a signal larger than 8 microns. All analyses were performed per mouse and for each mouse three slides were measured (Nguyen and Nguyen, 2013).

As tMCAo can cause changes in the posterior circulation as well (Lee et al., 2014), we examined the hippocampal formation and the ipsilateral hemisphere.

Cerebral spinal fluid and plasma collection in mice

Cerebral spinal fluid and plasma samples were collected from anaesthetized animals 15 days after tMCAo and Sham surgeries and prior to transcardial perfusion. Mice were anaesthetized via intraperitoneal injection with a mixture containing ketamine (100 mg/kg body weight) and xylazine (10 mg/kg body weight). For CSF samples, anaesthetized mice were placed in a prone position and the skin covering the back of the neck was shaved. A cotton swab containing 70% ethanol was used to remove any hair from the exposed skin. Then, a 27-gauge sterile needle (SV*27EL, Terumo Medical Products) attached to a 1-ml syringe (329650, BD Biosciences) was inserted into the cisterna magna allowing the flow of CSF into the butterfly needle. After 10–15 s, the needle was removed and the CSF aspirated into microcentrifuge tubes (1605-0000, USA Scientific), followed by a brief centrifugation at 600×g for 6 min at 4°C. The supernatant was transferred to a new tube, immediately placed on dry ice and further stored at −80°C. Roughly, 5–10 µl of fluid was collected per mouse. CSF visibly contaminated with blood (pelleted residual erythrocytes) was discarded. All remaining samples underwent a more stringent assessment for blood contamination via haemoglobin enzyme-linked immunosorbent assay (cat# ab157715, Abcam) using 0.5 µl CSF with a 1:200 dilution. Samples below 0.01% blood contamination were used for tau quantitation.

For plasma samples, whole blood collected from the submandibular vein was directly poured into anticoagulant Ethylenediamine tetraacetic acid-treated tubes. Animals were restrained with one-hand technique and a lancet (5 mm) was used to puncture the submandibular vein. Samples were centrifuged at 2000×g for 15 min at 4°C. The supernatant was transferred to a new tube, immediately frozen in liquid nitrogen and further stored at −80°C.

Cerebral spinal fluid and plasma tau quantitation in mice

Cerebral spinal fluid and plasma tau were quantified by Simoa technology using the murine total-tau assay (cat# 102209, Quanterix). Individual CSF and plasma samples were diluted in sample buffer at 1:75 and 1:4, respectively, and then split into technical duplicates (100 µl/replicate). Standards and samples were run according to the manufacturer’s instructions and read on the SR-X analyzer (Quanterix).

Statistical analysis

Demographic data, including age and sex, and biomarker data, including percentage of individuals classified as amyloid positive, plasma tau levels and total WMH volume, were compared across the three diagnostic groups (NC, MCI and Alzheimer’s disease) with analysis of variance for continuous variables and Chi-squared analysis for proportional data. For primary hypothesis testing, we used a mixed design general linear model that included diagnostic group, amyloid status (negative and positive), plasma tau levels, the interaction between diagnostic group and plasma tau levels, and the interaction between amyloid status and plasma tau levels as primary predictors of total WMH volume. Mean centred participant age was included in the model as a covariate. In the case of interactions, the data were stratified by the categorical variable (e.g. diagnostic group) and multiple regression analysis was used to test associations within group. We hypothesized that increased plasma tau levels would be associated with increased WMH volume, and that this relationship would be stronger across diagnostic groups (NC < MCI < Alzheimer’s disease). This model was repeated with CSF-derived measures of total tau and phosphorylated tau.

To test the possibility that observed effects involving WMH volume reflect associations with Alzheimer’s disease-related neurodegeneration in general (and not changes related to ischaemia per se), we re-ran the mixed design general linear model designed to test the primary hypothesis with relative total brain volume and relative white matter volume as dependent variables. We then repeated the primary analysis adjusting for relative total brain volume.

We used a series of binary logistic regression analyses to examine the extent to which amyloid positivity, plasma tau levels, WMH volume and the interaction between WMH volume and tau levels could classify participants into diagnostic groups. We ran separate analyses that contrasted Alzheimer’s disease participants to controls, Alzheimer’s disease participants to MCI participants and MCI participants to controls; these analyses included mean-centred age as an additional covariate.

For postmortem neuropathology validation, we performed a regression analysis, adjusting for age at death, to examine the relationship between the arteriosclerosis index and neurofibrillary tangle pathology.

In mice, plasma and CSF tau concentrations were compared between tMCAo mice and Sham controls with general linear models. We examined the relationship between plasma and CSF-derived concentrations with Pearson correlations. Similarly, RU and IF signals, reflecting tau and MBP levels, obtained from three slides per mouse were compared between tMCAo and sham mice with general linear models.

Data availability

ADNI data are available to the general scientific community through online request (http://adni.loni.usc.edu/data-samples/access-data/).

Results

Human data

Table 1 displays demographic and biomarker data for ADNI participants across the three diagnostic groups. The three groups were similar in age and sex distribution. Amyloid positivity and plasma tau levels varied reliably across groups in the expected direction: a greater proportion of individuals diagnosed with Alzheimer’s disease were classified as amyloid positive compared with those diagnosed with MCI, who in turn, were more likely to be amyloid positive than controls. Plasma tau levels increased monotonically across the three diagnostic groups.

Results from the primary mixed design general linear model that examined the relationship of plasma tau concentrations and WMH across diagnoses showed a trend for increasing WMH volume across diagnostic groups [F(2,382) = 2.98, P = 0.05; see Fig. 1]. Increased plasma tau concentrations were associated with higher WMH volume [main effect of plasma tau, F(1,382) = 5.27, P = 0.022, B = 0.743 [0.259, 1.228]; see Fig. 2], and this relationship differed across diagnostic groups [diagnostic group by plasma tau interaction, F(2,382) = 4.95, P = 0.008)]. The analysis stratified by diagnostic group confirmed that the magnitude of the relationship between plasma tau levels and WMH volume increased across the diagnostic groups and was strongest among participants diagnosed with Alzheimer’s disease (NC: B = −0.083[−0.34 to 0.17], MCI: B = 0.038[−0.15 to 0.22], Alzheimer’s disease: B = 0.448 [0.16–0.72]; see Fig. 3). Neither amyloid positivity [F(1,382) = 1.32, P = 0.25] nor the amyloid positivity by tau plasma concentration interaction [F(1,382) = 2.16, P = 0.14] was associated with WMH volume. Older age was associated with greater WMH volume (F(1,382) = 12.09, P = 0.03, B = 0.050 [0.022–0.079]). When the analyses were repeated with CSF total tau and phosphorylated tau rather than plasma levels, there was a trend for a positive association between WMH and the CSF tau measures [F(1,382) = 1.67, P = 0.19 and F(1,382) = 2.93, P = 0.09 for main effects of total and p-tau, respectively], but these relationships did not vary as a function of diagnostic group [F(2,382) = 0.37, P = 0.68 and F(2,382) = 0.19, P = 0.82 for total tau and p-tau by diagnostic group interactions, respectively].

Figure 1.

White matter hyperintensity (WMH) volume increases across diagnostic groups. Results from the primary mixed design general linear model shows increasing WMH volume across clinical diagnostic groups. Midline represents the median, the box denotes the 25th percentile and 75th percentile, T-bars represent 95% confidence intervals, points and asterisks represent outliers and extreme outliers, respectively. Y-axis was transformed logarithmically (base10) for visualization. AD = Alzheimer’s disease; NC = normal control; MCI = Mild Cognitive Impairment.

Figure 2.

Linear association between white matter hyperintensities (WMH) and plasma tau. Main effect of predicted WMH plotted against predicted plasma tau values obtained from the primary mixed design general linear model [F(1,382) = 5.27, P = 0.022, r² =0.153, β = 0.743 (95% CI 0.259–1.228)].

Figure 3.

Interaction of WMH and plasma tau levels across diagnostic groups. Stratification across diagnostic groups demonstrates interactions between white matter hyperintensities and plasma tau burden at different diagnostic stages. Participants with Alzheimer’s disease had the strongest association [NC: β = −0.083 (95% CI −0.340 to 0.175), MCI: β = 0.038 (95% CI −0.152 to 0.229), Alzheimer’s disease: r² = 0.677, β = 0.448 (95% CI 0.168 to 0.728)].

Plasma tau levels were not related to relative total brain volume as a main effect [F(1,382) = 0.731, P = 0.39] or interacting with diagnosis [F(2,382) = 1.31, P = 0.27] nor were they related to relative white matter volume as a main effect [F(1,382) = 0.255, P = 0.61] or interacting with diagnosis [F(2,382) = 0.221, P = 0.80]. When we re-ran the primary analysis, above, adjusting for relative total volume, we found the main effect of plasma tau [F(1,381) = 6.26, P = 0.01] and the interaction between plasma tau levels and diagnosis [F(2,381) = 4.72, P = 0.009] to be of similar or even greater magnitude.

Results of the logistic regression analyses that examined whether the biomarkers discriminate among diagnostic groups are presented in Table 2. In the analysis discriminating participants diagnosed with Alzheimer’s disease from controls, 77.6% of participants were correctly classified (χ² = 88.91, P < 0.001, sensitivity 89.7%; specificity 66.7%). Amyloid positivity was highly associated with an increased likelihood of being diagnosed with clinical Alzheimer’s disease. In this model, increased WMH volume was associated with a lower likelihood of being diagnosed with clinical Alzheimer’s disease, and plasma tau levels were not associated with diagnosis (i.e. main effects of WMH volume and plasma tau levels). However, without the interaction term in the model, WMH volume trended positively with Alzheimer’s disease classification. That is, including an interaction term with plasma tau (B = 0.49*tau) in the model (Table 2), effect of WMH on diagnosis was negative (B = −1.18), but the total effect of WMH (B = −1.18 + 0.49*tau) becomes positive at higher levels of plasma tau (above tau of 2.41, B > −1.18 + 0.49*2.41 > 0). Therefore, the total effect of WMH predicts that higher WMH volume, at higher levels of plasma tau, is associated with a greater likelihood of being Alzheimer’s disease over NC. Plasma tau levels and WMH interacted to increase the likelihood of Alzheimer’s disease diagnosis, such that the individuals with increased WMH volume were particularly likely to be classified as clinical Alzheimer’s disease when tau levels were also elevated. Very similar results were observed in the analysis that discriminated MCI participants form controls [χ² = 45.99, P < 0.001, 63.3% correctly classified (sensitivity 78.0%; specificity 61.1%)]; that is, increased amyloid, decreased WMH volume and the interaction between plasma tau levels and WMH volume were associated with a greater likelihood of MCI diagnosis. In the analysis discriminating Alzheimer’s disease participants from those with MCI [χ² = 21.807, P = 0.001, 66.1% correctly classified (sensitivity 4.1%; specificity 98.4%)], only amyloid positivity was associated with an increased likelihood for diagnosis of Alzheimer’s disease.

Table 2.

Results of a series of logistic regression analyses that discriminate between clinical diagnostic groups based on amyloid, tau, WMH and the interaction between WMH and tau

	AD (1) versus NC(0)			MCI (1) versus NC (0)			AD (1) versus MCI (0)
Variable	B	Odds ratio (95% CI)	P-value	B	Odds ratio (95% CI)	P-value	B	Odds ratio (95% CI)	P-value
Age	<0.001	1.00 (0.949–1.054)	0.995	−0.036	0.965 (0.927–1.004)	0.077	0.016	1.016 (0.981–1.053)	0.369
Amyloid positivity	3.04	20.91 (8.68–50.39)	<0.001	1.59	4.894 (2.893–8.280)	<0.001	1.518	4.565 (1.944–10.721)	<0.001
Plasma tau	0.03	1.03 (0.741–1.434)	0.855	−0.073	0.930 (0.778–1.112)	0.427	0.036	1.036 (0.872–1.231)	0.684
WMH volume	−1.18	0.307 (0.104–0.907)	0.033	−0.619	0.538 (0.309–0.938)	0.029	−0.172	0.842 (0.504–1.409)	0.513
Plasma tau x WMH volume	0.49	1.624 (1.049–2.514)	0.030	0.239	1.270 (1.029–1.567)	0.026	0.066	1.068 (0.908–1.255)	0.426

A 95% confidence interval (CI) was reported for each interaction.

AD = clinical Alzheimer’s disease; MCI = mild cognitive impairment; NC = cognitively normal controls; WMH = white matter hyperintensity.

Next, we turned to neuropathological data to confirm the association we observed between WMH and plasma tau concentration reflects a relationship between small vessel disease and tau pathology. Demographic data and descriptive neuropathological data for the autopsy sample are displayed in Table 3. When examining the postmortem data, we observed a positive relationship between arteriosclerosis severity and neurofibrillary tangle severity (B = 0.06, F = 4.21, P = 0.04).

Table 3.

Demographic characteristics, Thal and Braak stages and arteriosclerosis/atherosclerosis severity for autopsy cases included in the study

		NC	MCI	AD	Total	Statistic
N		7	33	23	63	-
Age, mean (SD) years		84.29 (6.78)	81.97 (6.15)	82.26 (8.73)	82.33 (7.18)	F = 0.30   P = 0.745
Sex, n (%) women		2 (28.6%)	6 (18.2%)	6 (26.1%)	14 (22.2%)	χ2 = 0.67   P = 0.714
Participants from primary analysis, n (%)		2 (7.4%)	17 (63.0%)	8 (29.6%)	27 (42.9%)	χ2 = 2.21   P = 0.332
Thal phase, n (%)	0–I	3 (42.9%)	5 (15.2%)	0 (0%)	8 (12.7%)	χ2 = 10.2   P = 0.037
	II–III	1 (14.3%)	2 (6.1 %)	2 (8.7%)	5 (7.9%)
	III–IV	3 (42.9%)	26 (78.8%)	21 (91.3%)	50 (79.4)
Braak stage, n (%)	0–I	2 (28.6%)	2 (6.1%)	0 (0%)	4 (6.3%)	χ2 = 15.1   P = 0.019
	II–III	2 (28.6%)	9 (27.3%)	2 (8.7%)	13 (20.6%)
	IV–V	3 (42.9%)	20 (60.6%)	15 (65.2%)	38 (60.3%)
	VI	0 (0%)	2 (6.1%)	6 (26.1%)	8 (12.7%)
Arteriosclerosis/atherosclerosis Index, mean (SD)		10.86 (6.26)	14.00 (6.86)	15.43 (7.66)	14.17 (7.12)	F = 1.13   P = 0.329
NIA-AA AD Neuropathologic Score n, (%)	Not AD	2 (28.6%)	1 (3.0%)	0 (0%)	3 (4.8%)	χ2 = 19.3   P = 0.013
	Low	1 (14.3 %)	10 (30.3%)	2 (8.7%)	13 (20.6%)
	Intermediate	1 (14.3%)	2 (6.1%)	0 (0%)	3 (4.8%)
	High	3 (42.9%)	19 (57.6%)	21 (91.3%)	43 (68.3%)

AD = Alzheimer’s disease; MCI = mild cognitive impairment; NC = cognitively normal controls NIA-AA = National Institute on Aging-Alzheimer’s Association.

Mouse model data

We turned to a mouse tMCAo model to confirm the correlations observed in human data and to test whether induced hypoperfusion causes ischaemic injury to white matter and associated tau pathology. Hypoperfusion via tMCAo produced focal infarct and widespread white matter ischaemic damage, based on MBP IF staining analysis. Mice exposed to 60 min tMCAo had lower IF MBP signal in the corpus callosum (307.41 ± 41), followed by mice exposed to the 30 min tMCAo (357.11 ± 42.3), and then Sham controls (428.5 ± 27; F(1,5) =18.05, P = 0.01). Similarly, IF MBP values were lower in the alveus of the CA1 subregion of tMCAo mice, with 60 min (283.2 ± 44.1), 30 min (355 ± 73.12) and Sham (451.6 ± 93.08; F(1,5) = 8.08, P = 0.04; see Figs 4 and 5).

Figure 4.

tMCAo produces MBP signal loss and induces tau phosphorylation. Representative microphotographs of MBP and AT8 IF in mice without stroke (sham, A and D), tMCAo for 30 min (B and E) and 60 min (C and F). tMCAo produced a reduction of MBP (red signal; see arrows in C and F) and induced tau phosphorylation tau (AT8; green signal; see arrows in C, F and inset in F). The degree of ischaemia-induced myelin loss and phosphorylation of tau increased with the increasing duration of ischaemia (tMCAo) from 30 to 60 min. Panels A–C shows double IF (MBP and AT8) with DAPI and panels D–F without DAPI. a = alveus of CA1; corpus callosum; p = stratum pyramidal CA1; o = stratum oriens of CA1; Scale bar = 50 μm.

Figure 5.

Mice subjected to tMCAo followed by 15 days reperfusion show strong alterations in the pattern and expression of MBP in corpus callosum and CA1 region of the hippocampus. Shown are three examples of mice subjected to 60 min occlusion surgery (A–C), which demonstrate a drastic change in MBP expression compared to sham surgery controls (D–F). In addition, tMCAo affects the MBP distribution pattern (compare A–C to D–F). These results suggest myelin loss and axonal damage due to ischemia-reperfusion that is still observed after 15 days. MBP colormap intensity pseudocolor-scale, red is the highest intensity and blue the lowest. a = alveus; cc = corpus callosum; o = stratum oriens; p = stratum pyramidale. Each panel represents an independent mouse (A–C; n = 3 and D-F; n = 3). Scale bar = 50 μm.

Using immunohistological analysis, we further observed that although Tau5 (pan–tau) immunoreactivity in the hippocampal regions did not differ greatly between 30 or 60 min tMCAo conditions (27.44 ± 7.5 RU versus 27.88 ± 1.9 RU), there was greater reactivity observed in both tMCAo groups than in Sham mice (20.52 ± 0.4; F(2,12) = 4.575 P = 0.02); formal post hoc testing showed, relative to Sham mice, a trend increase in the 30 min condition (P = 0.05) and a significant increase in the 60 min condition (P = 0.03; focal changes in staining can be seen Fig. 6). In addition, the pattern of staining produced by AT8 showed accumulation of abnormal phosphorylated tau in the CA3 region of the ipsilateral hippocampus in tMCAo compared with sham mice, which was particularly prominent in 60 min tMCAo mice (60.53 ± 1.6 RU) but also in 30 min tMCAo mice (47.43 ± 3.03 RU) when compared with Sham controls (22.8 ± 1.4 RU; F(2,12) = 26.8, P = 0.0001; post hoc testing showed reliable increases in both conditions (P = 0.0001) relative to Sham (see Fig. 6). Similarly, IF signal for AT8 evaluating cells labelled with signal > 8 µm in stratum pyramidal CA1 was highest in 60 min tMCAo mice (47.17 ± 7.2), followed by 30 min tMCAo mice (31.2 ± 19) and Sham mice (12.33 ± 6.7; F(1,5) = 18.21, P = 0.004; see Fig. 4).

Figure 6.

Mice undergoing tMCAo for 30 and 60 min followed by 15 days reperfusion exhibit higher total and phosphorylated tau immunostaining in the hippocampus and corpus callosum. Depicted here are representative images of total (Tau-5) and phosphorylated (AT8) tau immunostaining in the CA3 (A–C) and CA1 hippocampal subregions that includes white matter (alveus and corpus callosum areas) (D–F). Thirty minutes tMCAo produced a minor increase in Tau-5 and significant changes in AT8 staining in the stratum oriens (see asterisks in B and E) and corpus callosum (see arrows in E) compared to Sham (non-ischaemia surgery A and D) control. Sixty minutes tMCAo showed a greater increase of phosphorylated tau (AT8) immunostaining in pyramidal cells layer of CA3 (arrowheads in C) in addition to significant high intensity of total tau (Tau-5) in the CA1 pyramidal cell layer and in the corpus callosum (see arrowheads and arrows in F). a = alveus; cc = corpus callosum; o = stratum oriens; p = stratum pyramidale; r = stratum radiatum; WM, white matter; Scale bar = 100 μm.

Next, we investigated if tau changes observed in the brains of tMCAo animals and suggested by the human data, followed the exact same trend in CSF and plasma biofluids. For this purpose, Simoa technology was used for tau measurements, as described in Methods. Figure 7 shows tMCAo-associated differences in plasma and CSF tau levels across tMCAo and Sham groups. CSF and plasma total tau levels were elevated among mice following 60 min tMCAo (mean ± SD = 339.55 ± 136.08 pg/ml; mean ± SD= 183.34 ± 53.82 pg/ml, respectively) compared with mice following 30 min tMCAo (mean ± SD = 107.98 ± 10.92 pg/ml; mean ± SD= 50.37 ± 42.56 pg/ml, respectively) and Sham controls (mean ± SD = 138.37 ± 61.93 pg/ml; mean ± SD= 80.77 ± 26.81 pg/ml, respectively; F(2,11) = 7.38, P = 0.013 for CSF tau and F(2,11) = 10.70, P = 0.004 for plasma tau). Furthermore, plasma and CSF tau measurements were highly correlated with each other (R² = 0.522, β = 0.36, P = 0.008).

Figure 7.

CSF and plasma tau concentrations differ between mice exposed to tMCAo and sham controls. (A) CSF and (B) plasma tau concentrations were particularly pronounced for 60 min tMCAo [F(2,11) = 7.38, P = 0.013 and F(2,11) = 10.70, P = 0.004, respectively]. Data are expressed as mean ± S.E.M. (C) Plasma and CSF values were strongly correlated (R² = 0.522, β = 0.36, P = 0.008). n = 7 (Sham), n = 2 (tMCAo 30 min) and n = 3 (tMCAo 60 min). Sixty minute condition displayed in red, 30 min condition displayed in grey and sham controls in white.

Discussion

White matter hyperintensities are considered a radiological marker of small vessel CVD due to regional hypoperfusion (Wardlaw et al., 2013) and have been implicated in the clinical expression of Alzheimer’s disease (Brickman et al., 2015). Here, we demonstrate a direct relationship between increased WMH volume and plasma total tau levels, particularly among individuals diagnosed clinically with Alzheimer’s disease, suggesting the effect of small vessel CVD on risk and expression of Alzheimer’s disease is due partially to its promotion of tau pathology. The interaction between WMH and tau, independent of amyloid pathology, was associated with an increased likelihood of clinical Alzheimer’s disease and MCI, and postmortem markers of small vessel CVD were associated with severity of neurofibrillary pathology. Results from the human experiment are consistent with an emerging literature highlighting a tau pathway for Alzheimer’s disease pathogenesis that is independent of beta-amyloid pathology (Weigand et al., 2020); our findings suggest that cerebrovascular abnormalities may be one initiator of this pathway. The results from the human studies confirm a relationship between markers of tau and WMH of presumed vascular origin.

The conceptual model we were testing was based on the hypothesis that ischaemia or hypoperfusion damage that manifests radiologically as WMH causes or promotes Alzheimer’s disease-related tau pathology independent of beta-amyloid. Our previous work showed that baseline measures of WMH predict longitudinal increases in CSF tau, but not vice versa (Tosto et al., 2015), and recent studies show that forebrain hypoperfusion that occurs in hypoxia-ischaemia-induced white matter injury is associated with increased phosphorylated tau levels in rats (Raz et al., 2019). Because cross-sectional analyses with a human biomarker or postmortem data indicate co-dependency but not necessarily causality, we used a tMCAo mouse model to test the possibility that hypoperfusion results in white matter damage and can indeed induce tau pathology. Findings from this mouse model demonstrate that cerebral hypoperfusion results in decreased myelin (i.e. white matter ischaemic damage) and increased tau pathology, providing experimental support for the correlational analyses in humans and for our conceptual model. Only total-tau levels were available in the human data and previous work suggests that although total-tau levels are also elevated in the context of Alzheimer’s disease, phosphorylated tau plasma levels may be a more specific marker (Mielke et al., 2018); thus the observation of a clear relationship between induced hypoperfusion in mice and distributed phosphorylated tau levels provides evidence of a specific relationship between cerebrovascular injury and Alzheimer’s disease pathology. The abnormal tau phosphorylation observed following tMCAo mirrors what is seen in human Alzheimer’s disease. That is, the diagnostic antibody AT8 is specific to tau phosphorylation site S202/205, as seen in tauopathies such as Alzheimer’s disease (Braak et al., 2006). Further, in addition to observing increased tau phosphorylation throughout the ipsilateral occluded hemisphere, there was dramatically increased tau change in the ipsilateral hippocampal formation, where earliest tau-related changes are observed in human Alzheimer’s disease (de Calignon et al., 2012). Of note, in control mice, we did not observe robust tau signal in corpus callosum analysed (Fig. 6), which is likely due to the anterior location of the sections sampled, where normal tau is marginally seen compared with more medial or posterior sections and because normal tau has stronger labelling in non-myelinated than in myelinated axons (Kubo et al., 2019). We also performed identical Simoa-based measurements in mice and humans and confirmed increased CSF and plasma markers of tau pathology, as we observed in humans. We believe this study is among the first to consider Simoa-derived plasma tau biomarkers in murine models. Another unique aspect of our study is that, while most tMCAo studies use young-adult mice, in the present work, mice were aged to create a phenocopy more closely related to the human condition. Histological and biofluidic evidence of tau pathology in our study was more robust in the 60 min tMCAo condition than in the 30 min condition, suggesting a ‘dose effect’ of ischaemia on these markers. Our human-to-mouse study establishes the biological plausibility that ischaemia, known to manifest radiologically in humans as WMH, can indeed induce an Alzheimer’s disease-like tauopathy independent of beta-amyloid.

Together with others (Raz et al., 2019), our findings do not preclude the concomitant possibility that some degree of white matter damage results from tau-related neurodegeneration in a Wallerian-like fashion, a hypothesis tested by at least two other groups (McAleese et al., 2017; Graff-Radford et al., 2019). In one study (McAleese et al., 2017), cortical tau pathology, but not markers of CVD, correlated with white matter abnormalities in autopsied brain tissue derived from participants with Alzheimer’s disease, suggesting a neurodegenerative rather than a vascular source of the white matter changes. We suspect that in end-stage Alzheimer’s disease (i.e. at autopsy), in addition to parenchymal degeneration, there is degeneration of the small vessels, which may obscure relationships between vessel pathology and white matter abnormalities earlier in the disease in some autopsy series. In the current study, however, we confirmed that at postmortem markers of vessel disease (arteriosclerosis) are associated with neurofibrillary tangles. We additionally tested this possibility by adjusting for markers of degeneration (relative brain volume) and demonstrating that a reliable association between WMH and tau markers persists. The extant literature further highlights how midlife measures of hypertensive blood pressure are associated with postmortem markers of Alzheimer’s disease pathology alluding to chronic effects of hypoperfusion (Petrovitch et al., 2000), and that both antemortem and postmortem MRI-derived white matter abnormalities correlate with neurofibrillary pathology (Polvikoski et al., 2010; Brickman et al., 2018). Graff-Radford et al. (2019) showed no spatial or severity relationship between WMH volume and cortical tau pathology on positron emission tomography among community-dwelling older adults without dementia. Our findings of a lack of relationship between WMH and tau levels among individuals without dementia parallel these observations; in our analyses, the relationship was most prominent among symptomatic individuals, suggesting that ischaemic damage is driving some degree of tau pathology among those with impairment or that there is a critical threshold of ischaemic damage required to induce tau pathology and associated cognitive impairment. Finally, there is the possibility that the relationship between tau pathology and vascular abnormalities in humans may be bidirectional, as a recent study (Bennett et al., 2018) demonstrated there are morphological changes to blood vessels and increases in blood vessel density with the overexpression of tau.

According to current pathogenic theories of Alzheimer’s disease (Jack et al., 2013), tau accumulation with accompanying neurodegeneration and cognitive impairment comprises the final pathway to Alzheimer’s disease dementia. This study joins previous efforts in demonstrating an amyloid-independent effect of vascular abnormalities on this pathway; a recent study (Kim et al., 2018), for example, demonstrated that the relationship between small vessel CVD and cognitive status is mediated in part by tau pathology in the medial temporal lobe among patients with subcortical vascular impairment. These findings have both important theoretical and clinical implications. From a theoretical perspective, the possibility that small vessel CVD should be considered a core pathological feature of Alzheimer’s disease needs to be entertained, not simply because it confers additive risk to the clinical expression of the disease, but because it directly affects primary Alzheimer’s disease pathology. In this vein, the extent to which Alzheimer’s disease is diagnosed based on a single pathogenic pathway (Jack et al., 2018) should be questioned. From a clinical perspective, prevention of or intervention on vascular risk factors and CVD may have a direct effect on tau-related neurodegeneration and dementia and/or mitigate the pernicious effects of amyloid pathology.

Although previous studies demonstrated that cerebral hypoperfusion and ischaemia promote hyperphosphorylation of tau (Raz et al., 2019), the mechanisms linking the two are poorly understood. Compelling work with spontaneously hypertensive/stroke-prone rats showed increased blood brain barrier dysfunction following unilateral carotid artery occlusion, which in turn promoted neuroinflammatory white matter injury (Jalal et al., 2015). Similarly, in chronic hypertensive rats with induced hypoperfusion injury, there was white matter injury and an increase in endogenous phosphorylated tau coupled with increase in free radicals and active interleukin-1β (Raz et al., 2019). An increase in inflammatory microglia and the sensitivity of white matter mature and immature oligodendrocytes/precursors to ischaemia plays a role in white matter injury and associated cognitive decline in transient occlusion models (Nasrabady et al., 2018). Thus, we hypothesize that increased white matter injury due to ischaemia/hypoperfusion induces tau pathology via an inflammatory cascade. It is important to emphasize that this pathway is likely independent of amyloid pathology and points to hypoperfusion and inflammatory changes as potential targets for intervention or prevention.

We used a plasma-derived biomarker for total tau pathology, which has advantages and disadvantages. Plasma total tau tracks well with disease progression and is correlated reasonably well with central markers of tau derived from CSF and cognitive decline, but seems to operate independent of amyloid pathology (Mielke et al., 2017) and not considered accurate enough for diagnostic purposes (Mattsson et al., 2016). Plasma-derived markers of the phospho-tau181 isoform, however, may be more sensitive and specific to amyloid pathology than total tau (Mielke et al., 2018; Janelidze et al., 2020; Thijssen et al., 2020). When we examined CSF tau measures, our observations were in a similar direction but less statistically reliable, raising questions about whether the impact of CVD on tau biomarkers differs between CSF and blood markers. It is possible that neurons and/or glia within regions that are most affected by hypoperfusion are also more accessible to plasma membrane translocation, resulting in a higher concentration of tau secretion through plasma membrane (Brunello et al., 2019). Future studies will need to understand the factors that increase tau levels in the blood versus CSF. Only total tau concentrations were available to us in the current study. With evidence of elevated measures of phosphorylated-tau as a direct response to hypoperfusion in our mouse model, our findings contribute to the growing suspicion of a compounding pathology, suggesting that hypoperfusion induces white matter abnormalities that affect tau hyperphosphorylation. To further elucidate the relationship between hypoperfusion and tau elevation in Alzheimer’s disease, future studies will need to confirm that plasma-derived measures of phosphorylated tau are associated with markers of small vessel cerebral vascular disease.

In summary, our work suggests a codependency between WMH of presumed vascular origin and tau, which increases with clinical diagnosis of Alzheimer’s disease. We confirmed the biological plausibility that hypoperfusion–ischaemia induces tau pathology in a transient occlusion adult-mouse model. Future studies should focus on a better understanding of the causal relationships among CVD, inflammation and Alzheimer’s disease pathology in man and mouse.

Glossary

ADNI =

Alzheimer's Disease Neuroimaging Initiative

CSF =

cerebrospinal fluid

CVD =

cerebrovascular disease

MBP =

myelin basic protein

MCI =

mild cognitive impairment

MRI =

magnetic resonance imaging

NC =

normal control

NIA-AA =

National Institute on Aging-Alzheimer's Association

RU =

relative units

Simoa =

single molecule array

tMCAo =

transient middle cerebral artery occlusion

 

A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf