Background
Cerebral amyloid angiopathy (CAA) is characterized by amyloid deposition within the walls of leptomeningeal and cortical arterioles. Among the several types of amyloid proteins causing CAA, amyloid β (Aβ) is by far the most common. Aβ comprises several species of 39-43-residue peptides (including Aβ
1-40 and Aβ
1-42) that are produced from amyloid precursor protein (APP) via sequential proteolytic cleavage by β- and γ-secretases [
1‐
3]. Soluble Aβ monomers are produced throughout life; in certain individuals, these aggregate to form insoluble amyloid fibrils. This pathological form of Aβ is the major constituent of CAA. It is also the primary component of neuritic plaques - one of the pathological hallmarks of Alzheimer's disease (AD). The composition and pathogenesis of vascular vs. parenchymal amyloid deposits, however, have important differences. For example, while Aβ
1-42 is thought to be an important seed for the formation of both parenchymal plaques and CAA formation [
4,
5], higher Aβ
1-40 levels and increased Aβ
1-40/Aβ
1-42 ratios favor formation of CAA over parenchymal plaques in mouse models of AD [
6‐
9].
CAA is primarily a disease of the elderly, with about one-third of individuals aged 60 years or older demonstrating CAA upon postmortem histopathological examination. The incidence of CAA is even higher in patients with AD since these two conditions share common risk factors. Indeed, up to 90% of AD patients have histological evidence of amyloid deposits within cerebral vessels [
10,
11]. Clinically, CAA is a well-recognized cause of "lobar" hemorrhage in the elderly [
12,
13]. Several population-based autopsy studies indicate that CAA is also an independent risk factor for ischemic stroke and dementia [
14‐
18]. To further define the relationship between CAA and its neurological consequences, and to effectively examine novel therapeutics directed against CAA, definitive identification of CAA prior to patient death is critical. Yet, to date, definitive diagnosis of CAA is possible only by direct examination of pathological tissue. Short of obtaining such tissue via brain biopsy, only "possible" or "probable" diagnosis of CAA is achievable through use of the Boston Criteria, which utilize MRI to detect lobar microhemorrhage as an indirect indicator of CAA[
19]. This indirect diagnostic technique, however, is limited by its inability to quantify CAA severity and its reliance on cerebral hemorrhage as a surrogate marker for CAA[
19]. Development of a non-invasive method for selectively and accurately diagnosing and quantifying CAA would therefore be a major breakthrough for this disease.
Investigation into amyloid-imaging ligands for the diagnosis of AD and the evaluation of anti-amyloid therapy started more than 10 years ago [
20‐
24]. Fibrillar amyloid-binding dyes such as Congo red, chrysamine G, and thioflavins were investigated as ligands for positron emission tomography (PET) and single photon emission computed tomography (SPECT) imaging of amyloid deposits in AD patients. Utilizing radiolabeled forms of these molecules, however, was not clinically feasible due to their relative inability to cross the blood-brain barrier (BBB) and their low binding affinities for Aβ aggregates. Since the mid-1990s, many groups have attempted to develop CNS-accessible amyloid ligands derived from those molecules. To date, at least two amyloid tracers - [
11C]PIB ([
11C]6-OH-BTA-1) and [
18F]florbetapir ([[
18F]AV-45) - have been well characterized. Derived from thioflavin-T and styrylpyridine, respectively, both display favorable amyloid binding profiles, suggesting their great potential as a non-invasive method for early detection of AD and evaluation of anti-amyloid therapies in AD patients [
25‐
32]. However, neither dye is appropriate for the specific diagnosis and quantification of CAA due to their lack of selectivity for parenchymal versus cerebrovascular Aβ deposits as well as the low resolution of PET imaging.
During our laboratory's exploration into the effects of CAA deposits on neurovascular architecture and function in aged APP transgenic mice, we serendipitously observed that the fluorescent dye resorufin (7-hydroxy-3H-phenoxazin-3-one) appeared to selectively bind cerebral arterioles bearing congophilic fibrillar amyloid. In this study, we sought to further characterize this unique amyloid binding property of resorufin, and also explore the feasibility of utilizing resorufin and/or its derivatives for CAA-specific amyloid imaging.
Discussion
In the present study, we report three key findings: 1) that resorufin preferentially binds cerebrovascular Aβ deposits over neuritic plaques in aged Tg2576 mouse brains as well as in human AD brains; 2) that resorufin staining colocalizes to a congophilic dye methoxy-X34 in close proximity to dystrophic smooth muscle cells of CAA-affected vessels; and 3) that resorufin can be modified to enhance lipophilicity, while preserving marked selectivity for cerebrovascular Aβ deposits. These results indicate that the phenoxazine derivative resorufin and its derivatives are, to our knowledge, the first class of amyloid-imaging dyes that bind CAA in a highly selective manner. All previously described amyloid imaging ligands have been shown to bind CAA and neuritic plaques with similar affinity, making it very unlikely that these dyes could be used to develop PET imaging tracers appropriate for selective and definitive diagnosis of CAA in live patients. The unique selectivity of resorufin suggests that this class of dye has great potential as a CAA-specific amyloid tracer - the development of which would be a major diagnostic step forward for this frequent but often under-diagnosed condition.
Resorufin has been widely used as a fluorogenic probe to label bioactive molecules, and as an end-point product to measure hydrolytic activities of enzymes including peroxidases, cellulases, and aldehyde dehydrogenases. While examining the effect of CAA deposits on cerebrovascular oxidative stress, architecture, and function in aged Tg2576 mice, we initially observed that resorufin generated from Amplex red (a substrate for peroxidases) directly interacted with CAA independent of the status of oxidative stress in cerebral vessels. We characterized the cerebrovascular and parenchymal amyloid binding properties of resorufin and its derivatives ethoxy- and benzyloxy-resorufin. We found that the strongly fluorescent molecule resorufin preferentially bound cerebrovascular Aβ aggregates over neuritic plaques when
in situ staining was performed. In an independent study, Lebouvier et. al. [
37] have reported that resorufin binds to neuritic plaques, neurofibrillary tangles, and CAA in postmortem AD brain sections. However, the resorufin concentration used in that study was 2000-fold higher than that used herein (2 mM vs. 1 μM) [
37]. Non-selective binding of resorufin to Aβ is expected at a high concentration; importantly, however, their study did not examine whether lower concentrations of resorufin detect Aβ deposits differentially based on localization in cerebral vessels versus brain parenchyma. We observed markedly preferential binding to CAA when staining is performed with low concentrations of resorufin under stringent conditions (i.e., washing with PBS then with 50% ethanol-containing PBS). This preferential binding for CAA cannot be attributed to artifact during brain tissue processing or fluorescent labeling since resorufin selectively visualized CAA deposits when directly applied onto the cortical surface of live Tg2576 mice (Figure
4).
Two critical conclusions stem from our observations. First, fluorescent imaging with resorufin can be a highly useful tool for the selective histopathological evaluation of CAA. For example, it is currently difficult to quantify CAA versus neuritic plaque load using conventional amyloid dyes (e.g., thioflavin-S or Congo red analogs) due to their near equal affinity for vascular versus parenchymal Aβ deposits. By exploiting the preferential binding properties of resorufin for cerebrovascular amyloid plaques, this process of CAA quantification can be performed easily. Second, our data strongly suggest that the selective binding properties of resorufin can be exploited to eventually produce a CAA-specific amyloid tracer.
CAA is a strong and independent risk factor for cerebral hemorrhage, ischemic stroke and dementia in AD and non-AD patient populations [
12‐
18]. Excitingly, recent preclinical studies have identified several novel approaches that reduce or even prevent CAA formation [
38‐
41]. These studies raise the intriguing possibility that one or more of these CAA-directed therapeutic strategies might eventually be tested in humans. Unfortunately, such trials would currently be limited by the difficultly in diagnosing CAA: definitive diagnosis requires brain biopsy (which is rarely clinically indicated), and "probable" diagnosis of CAA by the Boston Criteria can be made only in patients who have already suffered intracerebral hemorrhage. Given that these bleeds occur less frequently [
19,
42] and at a later stage [
43] than ischemia, a trial using current diagnostics would be biased towards inclusion of later-stage CAA patients. The development of a non-invasive imaging technique for definitively diagnosing CAA would, in contrast, permit not only critical observational studies to better define the natural history of patients with CAA, but it would also greatly facilitate the organization and execution of therapeutic clinical trials that could include CAA patients who experience cerebral ischemia or dementia, not only hemorrhage.
To date, two chemically unrelated amyloid PET tracers, [
11C]PIB and [
18F]flobetapir, have demonstrated great promise as a tool for non-invasive amyloid imaging in patients with AD [
25‐
32]. Importantly, however, these PET tracers are unable to discern whether the observed amyloid load represents neuritic plaques or CAA since they label both parenchymal and cerebrovascular amyloid deposits [
44‐
46]. As such, our finding that resorufin analogs might represent a new class of PET agent for CAA-selective amyloid imaging is potentially groundbreaking. However, this must be considered in the context of well-described selection criteria for an ideal amyloid imaging PET tracer, including 1) high affinity and selectivity for target Aβ aggregates; 2) low molecular weight (< 400 g/mol); 3) moderate lipophilicity (logP
oct in a range of 1-3); and 4) functional groups amenable to labeling with a positron-emitting radionucleotide such as
11C or
18F - resorufin does not yet fulfill all of these requirements due to its low binding affinity for CAA (K
D: 874 nM) and low lipophilicity (logP
oct of 0.43) [
22]. Nevertheless, our pilot structure-activity relationship data show that chemical modification at resorufin's phenol group is able to improve binding affinity for CAA and increase lipophilicity while maintaining its high selectivity for cerebrovascular Aβ deposits. These results suggest that resorufin could serve as a lead compound to design chemical pools and to screen high-affinity, CAA-selective amyloid imaging dyes amenable to CAA imaging by PET.
Regarding the underlying mechanism by which resorufin analogs preferentially bind CAA over neuritic plaques, several potential explanations exist. One possibility is that resorufin binds fibrillar Aβ at different site(s) from other amyloid imaging dyes, a hypothesis that is supported by our observation that resorufin binding to CAA was not competitively inhibited by the congophilic dyes methoxy-X34 and methoxy-X04. A second possibility is that resorufin preferentially recognizes aggregations of Aβ
1-40 (the predominant species in CAA) over Aβ
1-42 (the predominant species in neuritic plaques). This hypothesis is supported by our observation that resorufin detects methoxy-X34-sensitive CAA from Tg2576 mice and humans (which is composed of both Aβ
1-40 and Aβ
1-42 [
6,
47]) (Figures
1 and
3), but does not detect methoxy-X34-sensitive CAA from BRI-Aβ42 transgenic mouse (which is composed almost exclusively of Aβ
1-42 [
5]) (data not shown). A third possibility is that resorufin directly interacts with molecules or proteins that are present in CAA but not in neuritic plaques. For example, heparan sulfate proteoglycans are expressed much more highly in cerebrovascular deposits as compared to neuritic plaques, both in human AD brains and in HCHWA-D mice carrying the Dutch-type amyloidosis [
48,
49]. Further investigation is required to elucidate the underlying mechanism by which resorufin preferentially binds CAA over neuritic plaques - the identification of which will not only shed new light on CAA pathophysiology but may also lead to novel therapeutic targets that could be exploited to help prevent CAA formation and its neurological consequences.
Competing interests
BHH, WC, RHM and GJZ have patent applications on the composition, methods, and use related to resorufin derivatives. The authors declare that they have no competing interests.
Authors' contributions
BHH contributed to the general administration and direction of the project, interpretation of experimental results, and development and writing of the manuscript. MLZ performed live fluorescent imaging. AKV performed multi-photon imaging in fixed brain sections and image processing, and writing of the manuscript. EM contributed to fluorescent imaging, statistical data analysis, and writing the manuscript. DHK performed the fluorescent ligand binding assays. JKG performed the CAA load and neuritic plaque load analyses. WC contributed to the experimental design for structure-activity relationship of test compounds. RHM contributed to the overall design of chemical modification and review of data. GJZ contributed to direction of the project, interpretation of experimental results, and critical reviewing of the manuscript. All authors read and approved the final manuscript.