Abstract
Introduction: Cerebral amyloid angiopathy (CAA) is characterized by amyloid β (Aβ) deposition in brain vessels, leading to hemorrhagic phenomena and cognitive impairment. Magnetic resonance imaging (MRI)-based criteria allow a diagnosis of probable CAA in vivo, but such a diagnosis cannot predict the eventual development of CAA. Methods: We conducted a retrospective cohort study of 464 patients with cognitive disorders whose data were included in a brain health biobank. De-identified parameters including sex, age, cognitive score, APOE status, and cerebrospinal fluid (CSF) levels of Aβ 1–40, Aβ 1–42, phosphorylated tau, and total tau were assessed in those with and without CAA. Odds ratios (ORs) and 95% confidence intervals (CIs) were determined. Results: CAA was present in 53 of 464 (11.5%) patients. P-tau level was significantly higher in those with CAA (115 vs. 84.3 pg/mL p = 0.038). In univariate analyses, the risk of developing CAA was higher with increased age (OR, 1.036; 95% CI: 1.008, 1.064; p = 0.011) and decreased CSF level of Aβ 1–40 (OR, 0.685; 95% CI: 0.534, 0.878; p = 0.003). In multivariate analyses, the risk of CAA remained higher with a decreased CSF level of Aβ 1–40 (OR, 0.681; 95% CI: 0.531, 0.874; p = 0.003). Conclusion: These findings suggest that Aβ 1–40 levels in the CSF might be a useful molecular biomarker of CAA in patients with dementia.
Introduction
Cerebral amyloid angiopathy (CAA) is characterized by the deposition of amyloid beta (Aβ) in the leptomeningeal and cortical blood vessels. This deposition is an age-dependent risk factor for intracerebral hemorrhage, ischemic stroke, and cognitive impairment [1]. The deposition of Aβ in the vessel wall can induce the release of inflammatory factors, complement activation, and oxidative stress, which can in turn lead to damage of vascular endothelium and smooth muscle cells, resulting in intracerebral hemorrhage and ischemic strokes and leading to cognitive decline [2].
At present, CAA can only be diagnosed via biopsy or postmortem examination [3, 4]. The development of magnetic resonance imaging (MRI) teqniques however allows the visualization of hemorrhagic phenomena in the brain, including signs of amyloid angiopathy [5]. The recently modified imaging-based Boston criteria for the diagnosis of CAA allow for a diagnosis of possible or probable CAA in vivo [6], although a definitive diagnosis still requires neuropathological assessment. The Boston criteria are based on two CAA-related imaging markers: lobar cerebral microbleeds (small brain bleeds restricted to the cortical and subcortical regions of the brain) and cortical superficial siderosis (deposition of blood products in the cortical sulci over the convexity of the cerebral hemispheres) [4].
Having the ability to recognize CAA in vivo has opened the door for studying its epidemiology, the genotypic and phenotypic spectra of the disease, and the existence of any potential biomarkers [7, 8]. Population-based clinical autopsy series and meta-analyses have demonstrated that CAA is a common neuropathological finding among individuals with various types of cognitive disorders [9, 10]. As a disease characterized by Aβ deposition, CAA is commonly compared to Alzheimer’s disease (AD). The manifestation of vascular (as in CAA) and parenchymal amyloid deposits (as in AD) can either overlap or occur independently. Both diseases are characterized by a pathologic amyloid metabolism, but the pathologic processing of amyloid proteins seems to be distinct in each condition [1, 9, 11]. A growing body of evidence suggests that cerebrovascular amyloid pathology is an independent contributor to cortical atrophy mediated by CAA-related vascular dysfunction [12], suggesting that CAA can cause cortical atrophy even in the absence of AD.
CAA is associated not only with the syndrome of cognitive impairment but also with immunotherapy-related side effects commonly seen in patients with AD. Currently, anti-amyloid monoclonal antibodies are emerging as a viable therapeutic option for individuals with mild cognitive impairment (MCI) and mild dementia due to AD. Passive immunotherapy lowers amyloid burden in the central nervous system but comes with a significant risk of amyloid-related imaging abnormalities (ARIAs). Although these abnormalities are usually asymptomatic and detected only on brain MRI, ARIAs can sometimes lead to new signs and symptoms such as headache, worsening confusion, dizziness, visual disturbances, nausea, and seizures. The risk of hemorrhagic ARIA increases with age and with the presence of cerebrovascular disease [13]. As such, current opinions are divided regarding whether the presence of imaging-recognized signs of CAA should disqualify patients from such treatment and which risk factors of amyloid angiopathy may predispose patients to the occurrence of hemorrhagic ARIA, especially as the lack of certain features of CAA on imaging (e.g., microbleeds) does not rule out the subsequent development of spontaneous microbleeds over time or of adverse outcomes [14, 15].
Our current understanding of the risk of CAA in patients with varying degrees of AD or in those with mixed pathology is limited. For instance, positive AD biomarkers do not rule out the presence of co-occurring secondary pathology. An awareness of the risk of CAA as a sole or coexisting secondary pathology is therefore important in allowing us to determine which patients could be candidates for novel immunotherapies when the risk for hemorrhagic ARIA is significant. Understanding the relationship between CAA and AD, and how they coexist and potentially interact, is crucial for accurate risk assessment and treatment decisions. This situation calls for the development of better biomarkers of CAA. In this study, we therefore sought to explore the relationships between imaging features of CAA, potential biomarkers, and risk factors of dementia in a cohort of patients with cognitive impairment.
Methods
The patients enrolled in this study were recruited from a specialized memory clinic at Cleveland Clinic (Lou Ruvo Center for Brain Health, Cleveland site). Collection of biomaterial was approved by the local Cleveland Clinic Institutional Review Board (14–604).
Individuals with CAA diagnosed based on validated imaging criteria [6] were identified among the 487 patients with various types of dementia included in the biobank. Patients with postsurgical implants, catheters, or the presence of intracranial air on MR images were excluded. De-identified demographic, genetic, clinical, laboratory, and cognitive data were then collected.
The diagnoses of MCI and AD were confirmed by the presence of cerebrospinal fluid (CSF) Aβ 1–42 and phosphorylated tau (P-tau) levels consistent with a diagnosis of AD as the primary etiology, as well as by a consensus evaluation conducted by two neurologists and based on published criteria [16]. Vascular dementia was diagnosed according to NINDS-AIREN criteria [17]. McKeith et al. [18] clinical criteria were followed for Lewy body dementia diagnosis.
Imaging
MRI scans were performed at 1.5 Tesla or 3.0 Tesla using Siemens scanners (Erlangen, Germany). All patients eligible for inclusion in this study had undergone both fluid-attenuated inversion recovery (FLAIR) imaging and susceptibility-weighted imaging (SWI); patients with only a gradient-recalled echo sequence were not included. The sequence parameters for FLAIR and SWI are shown in Table 1. A total of 464 individual records (images) were included in the final analysis. Five board-certified neuroradiologists, with experience ranging from 2 to 27 years, reviewed the images. Four of the neuroradiologists reviewed 100 images each, and the remaining neuroradiologist reviewed 64 images. The assessments for each study (including field strength, scanner manufacturer, and sequences performed) were entered into a RedCap database.
Sequence . | Scanner strength . | |
---|---|---|
1.5T . | 3.0T . | |
FLAIR | ||
Echo time, ms | 109 (89–160) | 130 (9–393) |
Repetition time, ms | 8,360 (7,270–9,000) | 8,770 (2,000–9,340) |
Inversion time, ms | 2,200 (2,000–2,500) | 2,500 (900–2,569) |
Slice thickness, mm | 4 (4–5.5) | 4 (1.0–4) |
In-plane voxel size, 9 mm | 0.82 (0.39–0.90) | 0.33 (0.33–1) |
SWI | ||
Echo time, ms | 40 (40–40) | 20 (20–20) |
Repetition time, ms | 49 (49–49) | 27 (26–27) |
Slice thickness, mm | 3 (2.0–4.0) | 3 (2.0–4.0) |
In-plane voxel size, 9 mm | 0.82 (0.82–0.90) | 0.82 (0.82–0.85) |
Sequence . | Scanner strength . | |
---|---|---|
1.5T . | 3.0T . | |
FLAIR | ||
Echo time, ms | 109 (89–160) | 130 (9–393) |
Repetition time, ms | 8,360 (7,270–9,000) | 8,770 (2,000–9,340) |
Inversion time, ms | 2,200 (2,000–2,500) | 2,500 (900–2,569) |
Slice thickness, mm | 4 (4–5.5) | 4 (1.0–4) |
In-plane voxel size, 9 mm | 0.82 (0.39–0.90) | 0.33 (0.33–1) |
SWI | ||
Echo time, ms | 40 (40–40) | 20 (20–20) |
Repetition time, ms | 49 (49–49) | 27 (26–27) |
Slice thickness, mm | 3 (2.0–4.0) | 3 (2.0–4.0) |
In-plane voxel size, 9 mm | 0.82 (0.82–0.90) | 0.82 (0.82–0.85) |
Values are means with values in parentheses representing the range of values across the various MRI scanners used.
SWI, susceptibility-weighted imaging; FLAIR, fluid-attenuated inversion recovery; T, tesla.
The number of microhemorrhages (entered as either a digit from 0 to 20 or as > 20) was separately recorded for the following locations (considered bilaterally): frontal lobe, parietal lobe, temporal lobe, occipital lobe, deep parenchyma, brainstem, and cerebellum. There were separate entries for pial siderosis and microhemorrhages sized >5 mm. The FLAIR sequence was assessed using the Fazekas score [19]. SWI foci within the basal ganglia and associated with lacunae were not counted.
Biomarkers
CSF biomarkers were measured using Luminex 200 × Map technology and the MILLIPLEX MAP® multiplex kit (EMD Millipore) following the manufacturer’s instructions for Aβ 1–40, Aβ 1–42, total tau (T-tau) and P-tau detection [20]. Ratios of CSF levels of Aβ 1–40 to 1–42 and well as Aβ 1–42 to 1–40 were calculated and included in the analysis.
Statistical Analysis
A Shapiro-Wilks test was used to assess normality for continuous variables. For normally distributed continuous variables, a t test was applied; for non-normally distributed continuous variables, a Wilcoxon rank sum test was applied. A χ2 test was used to compare categorical variables, and Fisher’s exact test was used to compare categorical variables. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for potential biomarkers of CAA. p values <0.05 were considered statistically significant.
Results
CAA was present in 53 of 464 (11.5%) patients with various types of dementia. In this analysis, 33% of those with CAA were diagnosed with AD; 38%, with vascular dementia; 23%, with MCI amnestic; 4%, with MCI non-amnestic; and 2%, with Lewy body dementia.
We evaluated the APOE polymorphism in our cohort of patients with cognitive impairment. At least one APOE4 allele was present in approximately 40.7% of the patients. Among those with CAA, the APOE4 allele was found in 43.5% of the cases. Conversely, the APOE2 allele is relatively rare in this context. It was present in only about 10% of patients with cognitive impairment but without CAA, and in around 13% of patients with both cognitive impairment and CAA (Table 2). P-tau level was significantly higher in those with CAA (115 vs. 84.3 pg/mL p = 0.038) (Table 2).
Variable . | All patients (N = 463) . | CI-CAA absent (n = 410) . | CI-CAA present (n = 53) . | p value . |
---|---|---|---|---|
Age | 66 (57–71) | 64 (57–71) | 69 (60–73) | 0.007 |
MoCA score | 21 (16–24) | 21 (16–25) | 19 (16–22) | 0.205 |
APOE 2 allele carriers, n (%) | 49 (10.6) | 42 (10.2) | 7 (13.2) | 0.081 |
APOE 4 allele carriers, n (%) | 190 (41.1) | 167 (40.73) | 23 (43.4) | 0.932 |
ATI | 0.47 (0.3–0.9) | 0.5 (0.3–0.9) | 0.4 (0.3–0.7) | 0.477 |
Sex, male, N (%) | 240 (51.84) | 215 (52.4) | 25 (47.17) | 0.652 |
CSF Aβ 1–40, pg/mL | 2,641 (1,803–3,832) | 2,730.5 (1,880.5–3,927.5) | 2,028 (1,010–3,290) | 0.001 |
CSF Aβ 1–42, pg/mL | 363 (185.5–551.5) | 370.3 (190.2–562.9) | 360(178–522) | 0.349 |
CSF Aβ 1–40/1-42 ratio | 8.2 (5.4–11.2) | 8.1 (5.5–10.9) | 8.841 (4.9–12) | 0.635 |
CSF Aβ 1–42/1-40 ratio | 0.14 (0.1–0.15) | 0.14(0.1–0.14) | 0.18(0.18–0.16) | 0.742 |
CSF T-tau, pg/mL | 552.7 (323–825) | 534.5 (308.5–808.7) | 605.8 (356–937) | 0.347 |
CSF P-tau, pg/mL | 89.6 (46–154) | 84.3 (41.9–149.1) | 115 (58–178) | 0.038 |
Variable . | All patients (N = 463) . | CI-CAA absent (n = 410) . | CI-CAA present (n = 53) . | p value . |
---|---|---|---|---|
Age | 66 (57–71) | 64 (57–71) | 69 (60–73) | 0.007 |
MoCA score | 21 (16–24) | 21 (16–25) | 19 (16–22) | 0.205 |
APOE 2 allele carriers, n (%) | 49 (10.6) | 42 (10.2) | 7 (13.2) | 0.081 |
APOE 4 allele carriers, n (%) | 190 (41.1) | 167 (40.73) | 23 (43.4) | 0.932 |
ATI | 0.47 (0.3–0.9) | 0.5 (0.3–0.9) | 0.4 (0.3–0.7) | 0.477 |
Sex, male, N (%) | 240 (51.84) | 215 (52.4) | 25 (47.17) | 0.652 |
CSF Aβ 1–40, pg/mL | 2,641 (1,803–3,832) | 2,730.5 (1,880.5–3,927.5) | 2,028 (1,010–3,290) | 0.001 |
CSF Aβ 1–42, pg/mL | 363 (185.5–551.5) | 370.3 (190.2–562.9) | 360(178–522) | 0.349 |
CSF Aβ 1–40/1-42 ratio | 8.2 (5.4–11.2) | 8.1 (5.5–10.9) | 8.841 (4.9–12) | 0.635 |
CSF Aβ 1–42/1-40 ratio | 0.14 (0.1–0.15) | 0.14(0.1–0.14) | 0.18(0.18–0.16) | 0.742 |
CSF T-tau, pg/mL | 552.7 (323–825) | 534.5 (308.5–808.7) | 605.8 (356–937) | 0.347 |
CSF P-tau, pg/mL | 89.6 (46–154) | 84.3 (41.9–149.1) | 115 (58–178) | 0.038 |
Numbers reflect mean values and range. All tests were two-tailed. p values <0.05 were considered statistically significant. In bold: statistically significant results.
CI, cognitive impairment; CAA, cerebral amyloid angiopathy; MoCA, Montreal Cognitive Assessment; CSF, cerebrospinal fluid; ATI, Amyloid to Tau Index.
Increased age and decreased CSF level of Aβ 1–40 were the only factors significantly associated with the occurrence of CAA, which typically manifested as microbleeds and/or cortical siderosis (Table 2). Ratios of CSF Aβ 1–40/1–42 and Aβ 1–42/1–40 did not differentiate between the groups with and without CAA (Table 2). In univariate analyses, the ORs were 1.036 (95% CI: 1.008, 1.064; p = 0.011) for age and 0.685 (95% CI: 0.534, 0.878; p = 0.003) for Aβ 1–40. In multivariate analyses, the ORs were 1.033 (95% CI: 0.992, 1.076; p = 0.121) for age and 0.681 (95% CI: 0.531, 0.874; p = 0.003) for Aβ 1–40.
Discussion
In this study of patients treated at a specialized memory clinic, we found that 11.5% of individuals demonstrated evidence of CAA on MR images. Decreased levels of Aβ 1–40 in the CSF were found to be associated with the occurrence of CAA.
CAA is the accumulation of amyloidogenic proteins in cerebral blood vessel walls, leading to a major risk for hemorrhage [21, 22]. Several types of hereditary disorders that result in CAA are caused by missense mutations within the Aβ precursor protein gene [22]. However, CAA most frequently occurs sporadically [23]. Previous research has demonstrated a 5%–7% prevalence of CAA in cognitively normal elderly individuals and 50%–57% in individuals with lobar intracerebral hemorrhage [23]. In our study, CAA was seen in 11.5% of patients with various forms of dementia, including AD and MCI, and the risk of CAA seemed to increase with age. However, we studied a cross-sectional cohort of patients, not a pathological sample. As has been shown previously, lack of CAA features on imaging does not rule out the subsequent development of CAA over time [14, 15].
Castellani et al. [11] postulated that amyloid accumulation represents a response to chronic stress and that the neurodegenerative process occurs at the neuronal level, encompassing aberrant cell cycle. Symptoms of CAA could therefore be conceptualized as an abrupt perturbation of vascular tissue homeostasis [11]. This hypothesis explains the pathology behind not only CAA but also ARIAs related to anti-amyloid monoclonal antibody administration, as well as the occurrence of hemorrhage after Aβ immunization.
Previous research has shown that this perivascular drainage of soluble Aβ may be affected by the presence of the APOE4 allele, which is a genetic risk factor shared by both CAA and AD [24]. The prevalence of the APOE4 allele in the general population varies by ethnicity, typically ranging from 13 to 15%. In our cohort, the prevalence was significantly higher, with at least one APOE4 allele present in approximately 40.7% of the patients. Among those with CAA, the APOE4 allele was found in 43.5% of cases. This elevated prevalence likely contributes to the observed cognitive impairment and CAA in our cohort. Levels of APOE protein in the cerebral vascular wall have been shown to increase after anti-Aβ immunotherapy [25], and research has demonstrated that APOE and Aβ may be co-localized in the perivascular drainage route [26]. After intraventricular injection of Aβ 1–40 in APOE transgenic mice, researchers observed co-deposition of APOE4 and Aβ 1–40 in the vessel wall rather than APOE3 [24], indicating that the drainage rate of Aβ 1–40 mediated by APOE4 is much slower. This suggests impaired clearance of the APOE4-Aβ complex through perivascular drainage when compared with drainage of other complexes [26, 27]. APOE4 has lower antioxidant activity than other APOE isoforms [28], thus accelerating the loss of vascular integrity and breakdown of the blood-brain barrier, both of which contribute to the development of CAA. Indeed, the APOE4 allele has been identified as a risk factor for both sporadic and AD-related CAA [29], whereas the APOE2 allele, normally “protective” against the development of AD, is associated with an increased risk of CAA-related hemorrhage [30]. It is possible that APOE2, while protective against AD, may have a different effect on blood vessel integrity or amyloid deposition in the brain, making individuals with this allele more susceptible to CAA-related hemorrhage. In the population of this study, a link between APOE2 and intracranial hemorrhage was not observed. We did not find a significant association between the APOE2 allele and intracranial hemorrhage, despite the established link between APOE2 and CAA-related hemorrhage [24].
In dementia research, amyloid and tau proteins are the most established and rigorously investigated biomarkers. The use of CSF Aβ 1–40 and 1–42, T-tau, and P-tau as biomarkers has been included in the new consensus research diagnostic criteria for AD, MCI, and preclinical AD [31]. Because of its hydrophobic nature, Aβ 1–42 has the propensity to form aggregates and oligomers, which in form fibrils that accumulate into amyloid plaques. Aβ 1–40 is less prone to aggregation compared to Aβ 1–42 but still contributes to the formation of amyloid plaques. Aβ 1–40 is generally considered less toxic than Aβ 1–42, but it can still contribute to AD pathology. Tissue content of Aβ 1–40 however does not correlate with senile plaques but with CAA [32]. Verbeek et al. [33], observed strongly decreased CSF concentration of both Aβ 1–40 and Aβ 1–42 in patients with CAA. In the current study, we found that lower CSF levels of Aβ 1–40 (as opposed to Aβ 1–42) were associated with the presence of CAA. This suggests that Aβ 1–40 deposition in vessels makes them vulnerable and leaky, eventually leading to hemorrhage that is recognizable on imaging. In the study of a relatively large group of patients, comparing those with possible CAA versus probable CAA, Aβ 1–40 and tau levels did not differ significantly between the two groups. However, Aβ 1–42 levels and the Aβ 1–40/1–42 ratio were lower in patients with probable CAA. This may suggest that the deposition of Aβ 1–40 may precede the deposition of Aβ 1–42 in the process of CAA development.
Pathological changes associated with dementia development are also reflected by an increase in the CSF concentrations of T-tau and P-tau. Increased CSF P-tau is intricately intertwined with disease severity and its association with the heightened risk of ARIAs [13‒15]. Our investigation revealed a conspicuous elevation of P-tau levels in individuals affected by CAA.
Notably, this heightened P-tau manifestation did not parallel the observations made in total tau levels, emphasizing the unique role of P-tau in the intricacies of CAA. The elevation of P-tau in those with CAA may serve as an indicator of a more profound and severe manifestation of the underlying pathology. There have been however conflicting findings when applying tau levels to CAA [21, 32].
In recent study of probable CAA and AD, CAA presented with elevated P-tau. Compared to AD, however, CAA showed lower levels of P-tau, and Aβ 1–40 but similar Aβ 1–42 level. The reasons for these discrepancies could be due to, the unique characteristics of CAA in different phases of its development [34] The reasons for these discrepancies could be due to, the unique characteristics of CAA in different phases of its development In CAA, P-tau worsens Aβ deposition in blood vessels, triggers inflammation, disrupts the blood-brain barrier, and hinders Aβ clearance, all contributing to amyloid angiopathy. Understanding these pathways is vital for developing treatments targeting tau-related pathology in CAA progression [14, 15, 32, 34, 35].
Limitations of the Study
The data were obtained from a single-center biobank. Some relationships or associations may not reach statistical significance due to the limited number of participants. Due to the data available and the sample size, we were unable to estimate a specific cutoff value for CSF Aβ 1–40 that could effectively distinguish between patients with CAA and those without CAA. Future research with larger and more diverse cohorts may provide valuable insights in this regard.
Conclusions
Our data align with established evidence, reaffirming that CAA as defined by the Boston criteria, exhibits correlation with diminished cerebrospinal fluid (Aβ 1–40 levels. These results emphasize the potential utility of Aβ 1–40 levels in the CSF as a valuable molecular biomarker for CAA in individuals afflicted by cognitive impairment.
Furthermore, our investigation underscores the intricate relationships between low Aβ 1–40, elevated CSF P-tau, and age in the context of CAA. These findings bear profound relevance for the formulation of therapeutic strategies, particularly in an era characterized by the emergence of monoclonal antibody-based treatments for neurodegenerative disorders.
Statement of Ethics
This study protocol was reviewed and approved by the Institutional Review Board at Cleveland Clinic, Approval No. 18-1084. The analyses were based on de-identified data from the Center for Brain Health Biobank. Written informed consent was not required as per Cleveland Clinic IRB.
Conflict of Interest Statement
Lynn Bekris is a director of the Center for Brain Health Biobank. James B. Leverenz is a director of the Cleveland Alzheimer’s Disease Research Center and a director of the Coordinating Center for Dementia with Lewy Body Consortium. He also serves on the advisory board of Vaxxinity Inc. The other authors have no conflicts of interest to declare.
Funding Sources
Samples were analyzed under Center for Brain Health Biobank funding (Jane and Lee Seidman Fund). Lynn Bekris received research funding from the National Institutes of Health. James B. Leverenz received research funding from the National Institutes of Health (CADRC P30AG072959), GE Healthcare, and the Lewy Body Dementia Association (DLBC U01NS100610).
Author Contributions
Kasia Gustaw Rothenberg designed the study, interpreted the data, and wrote the manuscript. Lynn Bekris selected data from the biobank and performed the biomarker assessment. James B. Leverenz interpreted the data and edited the manuscript. Jenny Wu, Jonathan Lee, Volodymyr Statsevych, and Paul Ruggieri analyzed and interpreted the MRI results. Stephen Jones analyzed and interpreted the MRI results and edited the manuscript.
Data Availability Statement
All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.