Targeting β-amyloid plaques and oligomers: development of near-IR fluorescence imaging probes

Abstract

Evidence indicated that shifting treatment to a presymptomatic stage may produce significant benefits to prevent/alleviate the progression of Alzheimer's disease (AD); in particular, early incorporation of noninvasive imaging and biomarker testing will be significantly beneficial for AD drug development. Based on amyloid cascade hypothesis and its revised version, both β-amyloid deposition and soluble oligomeric species could be good diagnostic biomarkers for AD. Near-IR fluorescence (NIRF) imaging, which so far is limited to animal studies, is a promising method for its incomparable advantages such as low cost, high-throughput and easy operation. This review focuses on recent reported NIRF probes that showed excellent binding to plaques and oligomers. We hope that this review will shed light on the future of NIRF probes’ discovery.

As the most common cause of dementia, Alzheimer's disease (AD) is posing a serious threat to public health and healthcare systems in both developed and developing countries. AD is a progressive neurodegenerative disorder with cognitive impairment symptoms, including memory loss, language barrier, odd behavior, personality and mood changes. By the age of 65 years old, about 1% of the population has AD, and it will likely increase to 50% in people over the age of 85 years old [1]. Due to the globally aging trends, the number of people suffering from dementia is ever growing [2,3]. Moreover, the financial cost of AD is forecasted to grow rapidly with the aging trends [4]. In the USA, an estimated 17.9 billion hours of care for AD patients and other dementias were provided by 15 million family members and other unpaid caregivers in 2014, which is valued >US$217 billion [5]. Clearly, AD has become one of the greatest healthcare challenges in the 21st century.

It has been >100 years since the first case report of AD was published by psychiatrist Alois Alzheimer in 1906 [6,7], AD is still an insurmountable challenge for physicians and medical researchers. There is no efficient therapy for preventing AD. Six drugs (donepezil [Aricept^®, Eisai Co Ltd, Tokyo, Japan]; rivastigmine [Exelon^®, Novartis Pharma Ltd, Basel, Switzerland]; galantamine [Reminyl^®, Johnson & Johnson, NJ, USA]; tacrine [Cognex^®, Pfizer, NY, USA]; memantine [Namenda^®, Merz Pharm, Frankfurt, Germany]; and [Namzaric^®, Adamas Pharmaceuticals, CA, USA) have been approved by the US FDA [6–8]. These approved drugs only provide symptomatic relief and short-term benefits, but they inadequately alleviate the progression of the disease. In fact, it is widely believed that severe pathological changes are already presenting in the brains of AD patients before the diagnosis becomes clinically apparent [9,10]. Previous research has shown that cerebrospinal fluid levels of Aβ_1–42 are fully changed 5–10 years before the onset of AD [11]. In humans, abnormal levels of β-amyloids (Aβs) in brain appear 30 years before the symptom starts. However, the current positron emission tomography (PET) probes can only detect the abnormal Aβ deposits around 5 years before the clinical syndrome, which is obviously too late for early diagnosis.

There still is no general agreement about the pathogenesis of AD, but four main hypotheses have been intensively studied: amyloid cascade hypothesis, tau hypothesis, metal ion hypothesis and oxidative stress hypothesis (Figure 1). The amyloid cascade hypothesis states that Aβ species play key roles in the pathological progression of AD [12–14]; the tau hypothesis underlines that hyperphosphorylation and subsequent mislocalization of tau protein are identified as seminal steps for AD pathogenesis [15]; the metal ion hypothesis emphasizes that the impaired metal homeostasis likely contributes to the pathology, in particular of abnormal high concentrations of Zn, Cu and Fe are the underlying causes of AD. The impaired metal homeostasis also leads to the over-accumulation of Aβs [16]; the oxidative stress hypothesis speculates involvement of the production of free radicals, which can influence metabolism and also promote Aβ aggregation [17]. Though various hypotheses focus on different features of the disease, as shown in Figure 1, these hypotheses all stress that Aβ species are involved in the physiological symptoms and play a vital role in the pathological process of AD [18].

Figure 1. . An overview of four main hypotheses of Alzheimer's disease.

Aβ: β-amyloid; APP: Amyloid precursor protein; FAD: Familial Alzheimer disease; GSK3β: Glycogen synthase kinase 3β; HIF: Hypoxia-inducible factor; PSEN₁: Presenilin 1; PSEN₂: Presenilin 2; ROS: Reactive oxygen species.

Data taken from [19–21].

The studies on the amyloid cascade hypothesis are an ongoing process. Initial hypothesis suggested that Aβ plaque in the brain was a central role in AD pathology, and this assumption had been the framework for research in the past 20 years. However, all of the Aβ-centric therapeutics that reached Phase III clinical trials failed (Bapineuzumab, Pfizer; Solanezumab, Eli Lilly and Company, etc.) [22–25], which led to question the roles of Aβ and amyloid deposition in AD pathology. Many researchers directed their attentions to tau protein, and considered that abnormal hyperphosphorylation of tau protein might be a key component in neurodegenerative processes [26–30]. Meanwhile, a redefinition of the amyloid cascade hypothesis was supported by increasing evidence from the literatures, in which smaller, soluble oligomeric species of Aβ were considered to contribute to either neuronal death and/or affect synaptic neurotransmission [19,31–34]. In addition, the results of Dominantly Inherited Alzheimer Network indicated AD mutation carriers exhibited high levels of CSF Aβ₄₂ at least 30 years before the estimated symptom onset, but CSF tau levels were increased approximately 15 years before the onset [35], which implicated the aggregation of Aβs might act as a critical early trigger in the chain of events [36,37]. Therefore, Aβ species, especially soluble Aβ, appear to be good diagnostic biomarkers for AD, and also be good predictive biomarkers of progression of AD [38].

Research progress in noninvasive molecular imaging probes

Different molecular imaging methods, including MRI, PET, single-photon emission computed tomography (SPECT) and optical imaging, have been used in attempt to detect the progression of AD through visualization of Aβs both in vitro and in vivo [39].

Current modalities of molecular imaging applied in clinical studies

MRI and PET are established imaging techniques for clinical investigations of AD with sensitivity and specificity reaching 85–90% [40]. However, on account of the low sensitivity, blurred signal contrast between Aβ plaques and surrounding tissues, and low blood–brain barrier (BBB) permeation of contrast agents, MRI is not ideal for molecular imaging of AD [41].

PET imaging is the most promising imaging modality for AD. Three ¹⁸F-labeled ligands have been approved for clinical use by the European Medicines Agency and the FDA: Florbetapir (Amyvid™, Eli Lilly), Florbetaben (Neuraceq™, Piramal Imaging Limited), and Flutemetamol (GE Healthcare) (Table 1) [42,43]. Owing to the excellent sensitivity of 10^-10-10^-12 mol/L, limitless tissues’ depth penetration and safety for biological material, PET is a high-performance tool used in clinical for AD [44]. Nonetheless, the expensive facilities and risk of radiation exposing, and short half-life of the tracers are inevitable limitations.

Table 1. . A brief introduction of positron emission tomography tracers for clinical application.

Characteristic	Flutemetamol	Florbetapir	Florbetaben
Synonyms	GE-067,3′-fluoro-PIB	AV-45	BAY-94–9172, AV-1
Amyloid affinity (Ki, nM)	0.7	2.2	2.4
Plasma metabolites	Polar	Polar and nonpolar	Polar and nonpolar
Typical injected dose (MBp)	185	300	300
Typical imaging time (min)	80–100	50–70	90–130
Effective radiation dose (mSv; μSv/MBq)	6.3 (33.8)	5.8 (19.3)	4.4 (14.7)

Data taken from [42].

SPECT, which is based on single-photon-emitting radioisotopes, is other form of nuclear imaging. The cost of SPECT is significantly less expensive than that of PET scans, partly because the nuclides commonly used in SPECT have a longer half-life (¹²³I, t_1/2 = 13.2 h; ^99mTc, t_1/2 = 6.02 h) and agents are relatively easily obtained than PET tracers. Currently, ^[123I]IMPY has been selected for human studies [45,46]. However, because SPECT collimator absorbs most of the photons, the sensitivity and spatial resolution of SPECT are lower than that of PET imaging [47].

In short, molecular imaging technologies that have been currently applied to image Aβs can not completely fulfill the needs of both preclinical and clinical diagnosis.

An alternative plan: near-IR fluorescence imaging

In the last decades, fluorescent imaging has attracted wide attentions due to their broad applications coupled with high sensitivity and specific detection methods [48]. Compared with PET and SPECT imaging modalities, near-IR fluorescence (NIRF) imaging has many advantages, such as high sensitive, safe detection without radiation, and moderate cost. In addition, the fluorochromes in the near IR (NIR) window have minimal autofluorescence from cellular or tissue components. Though NIRF imaging is so far limited to animal studies, it is an attractive tool for early AD detection in preclinical studies because of its excellent features. Moreover, the proper NIRF probes can be easily modified to be PET or SPECT ligands with incorporating radionuclides. These dual-modal probes are favorable for multimodality imaging by merging nuclear imaging with optical imaging, which could be important complementarily for current imaging methods.

Herein, we provide a review of reported NIRF probes that bind various Aβs, including insoluble Aβs, such as plaque, and soluble Aβs, such as oligomers. Within this context, we summarized different types of compounds in a chronological order, including the strategies of structural modification and the optical properties of representative probes. We hope that our review can shed light on the future of NIRF probe discovery.

The progress in near-IR fluorescence imaging

The standards of small compounds as NIRF probes for Aβs

A good NIRF probe for Aβs in the brain should have the following prerequisites to be fulfilled: high selectivity and affinity for Aβs (K_d <20 nM); high brain penetration and fast washout kinetics from normal brain regions; high in vivo metabolic stability; low-affinity binding with bovine serum albumin; suitable emission wavelength between 650 and 900 nm, a large Stokes shift and high quantum yield (QY); fluorescence properties change upon binding to Aβs (i.e., fluorescence intensity, fluorescence lifetime, emission wavelength and QY); relatively easy of synthesis [49]. These requirements set the standards for us to evaluate the reported NIRF probes.

NIRF probes for imaging Aβ plaques

Based on amyloid cascade hypothesis, extracellular accumulation of Aβ peptides into plaques plays a vital role in AD onset [12]. Though the correlation between AD severity and the Aβ accumulation plaques is not linear, in vivo detection of Aβ plaques still is a way to predict the progression of AD and provides an accurate index for Aβ-related treatments. To date, substantial research has been dedicated to developing NIRF probes binding to Aβ plaques [50,51].

Fluorescent probes derivatived from styryl dyes

As a class of fluorescent probes, styryl dyes are used to serve as mitochondrial labeling agents and membrane voltage-sensitive probes of cellular structure and function [52,53]. Chang and coworkers established a library of combinatorial wide-color range fluorescent styryl dyes of 320 compounds [54,55]. From screening of the library, 2C40 and 2E10 were found to be the best candidates for imaging of the amyloid plaques in brain slices (1–2, Figure 2). However, the wavelength of excitation/emission of 2C40 and 2E10 are far from the NIR region as shown in Table 2. Meanwhile, the presence of the charged moieties limits their penetrability of BBB. Thus, these two candidates are only suitable for staining plaques in vitro.

Figure 2. . The structural optimization of styryl dyes and the chemical structures of 2C40, 2E10.

Compared with 2C40 and 2E10, STB-8 is a new type of styryl dye that its free amino group is removed.

Table 2. . Binding and optical properties of the near-IR fluorescence probes based on scaffolds 1–3 reported in the early exploration stage.

No.	Scaffold	Name	MW	λex/λem	Affinity†	Φ (%)	Ref.
1	1	2C40	491.65	505/590‡	–	–	[55]
2	1	2E10	381.50	464/503‡	–	–	[55]
3	1	STB-8	383.15	373/407§	3.2 μM	–	[56]
4	2	AOI987	411.16	650/670¶	0.2 μM	61/41#	[57]
5	2	ASG236	427.22	665/695¶	–	28/13#	[57]
6	2	ASG237	409.16	658/677¶	–	61/#	[57]
7	2	AMQ987	423.21	–	–	–	[57]
8	3	NIAD-4	334.41	475/612††	10 nM	0.008/5‡‡	[58]
9	3	NIAD-11	400.47	545/690 ††	–	11‡‡	[59]
10	3	NIAD-16	361.48	470/720 ††	–	–	[59]

^†Binding constant to β-amyloid aggregates.

^‡λex and λem measured in aqueous HCl (pH 1.4) recorded with a Hitachi F-2500 FL spectrophotometer.

^§λmax (nm) upon binding to β-amyloid fibrils determined by use of a Hitachi F-2500 FL spectrophotometer in 100 mg/ml fibril in PBS solution (1 mM EDTA, pH 7.4).

^¶λex and λem measured in mouse serum.

^#QYs of compounds in methanol and mouse serum.

^††λex and λem measured in combination of 10 μM aggregated Aβ₄₀ in PBS and fluorophores at 5 μM concentration.

^‡‡QYs of compounds were obtained with or without aggregated Aβ₄₀; the standard of QY calculated for NIAD-4 was aqueous rhodamine 6G (Φ = 0.76) and standard for NIAD-11/16 was methanol solution of methylene blue (Φ = 0.03).

λem: Maximum emission wavelength; λex: Maximum excitation wavelength; MW: Molecular weight; QY: Quantum yield.

In order to increase the hydrophobicity of the compounds for better permeability of BBB, a series of styryl-based neutral derivatives of 2C40 and 2E10 were reported by same workgroup in 2007 via removing both the pyridinium positive charge and the free amino group [56]. On the basis of structure of 2C40, the charged moiety is replaced by styrylpyridine and styrylbenzene respectively; for 2E10, the unique quinaldine-styryl structure is retained. They found that STB-8 (3, Figure 2) could successfully penetrate BBB and specifically binds to Aβ plaques in ex vivo studies. However, these styryl-based neutral derivatives can not be used for in vivo imaging because their excitation and emission are still outside of NIR region. In addition, they have low affinities for Aβ plaques.

NIRF probes derivatived from oxazine dyes

Based on the scaffold of oxazine dye (Scaffold 2), Hintersteiner et al. reported a series of NIRF probes for senile plaque imaging (4–7, Figure 3) [57,60]. The compound A0I987 (4, Figure 3) has excellent optical properties, such as maximum absorption/emission in the NIR domain (λ ex = 650nm, λ em = 670 nm) and high QY (ϕ = 41%), that are favorable for in vivo imaging. The long wavelengths of absorption/emission owe to the structure of benzo[1,4]oxazin and the high value of QY benefits from the rigid and planar conformation. Both these two characters are necessary for in vivo imaging. It is noteworthy that AIO987 is a charged molecule, which likely has low BBB permeability. Nevertheless, as reported, the maximal plasma concentration of AOI987 was achieved within 15 min in brain after administration, which indicates that significant amount penetration of AOI987 passed through the intact BBB [57]. The fluorescent properties of AOI987 are very attractive and meet most of the requirements for NIRF probe for the detection of Aβ plaques both in vitro and in vivo. However, its moderate affinity (K_d = 220 nM) to Aβ aggregates indicates that further structure optimization is needed, and its small Stokes shift (25 nm) may lower the detection sensitivity [61].

Figure 3. . The synthesis scheme of novel oxazine dyes.

Compound AOI987 has optimal optical properties for in vivo imaging (maximum excitation wavelength = 650 nm, minimum excitation wavelength = 670 nm; quantum yield = 41%).

NIRF probes derivatived from 2, 2′-bithiophene

Compared with the complex structures of compounds that were reported in early exploration stage, a series of 2, 2′-bithiophene compounds NIADs (8–10, Figure 4) possess relatively simple structures and excellent features.

Figure 4. . Structural model of nonsteriodal anti-inflammatory drug and three typical compounds with sufficient research.

The design philosophy of NIADs roots in Congo Red (CR) and Thioflavin T (Th-T) that have planarized or easily planarizable π system [62–65]. Small planar structures that match with the features in amyloid fibrils surface are in part responsible for the high-binding selectivity to aggregated amyloids. With a classical push–pull architecture, terminal donor and acceptor moieties are interconnected by a highly polarizable bridge (Figure 4). Various donor and acceptor groups can be used to manipulate the relative energy gaps of highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO). Therefore the appropriate groups can lead to a smaller HOMO–LUMO gap, which can lead to the desired long excitation/emission wavelengths [66]. In addition, with the difference from derivatives of styryl and oxazine, NIAD probes bear no charged groups, and have high BBB permeability. To avoid high lipophilicity that will lead to its aggregation and accumulation in blood proteins and red blood cell membranes, NIADs possess some hydrophilic groups, such as thiophene and hydroxyl, to balance BBB permeability and affinity to Aβ fibrils. More interestingly, the significant increase of fluorescence QY of the bound probe indicates that NIADs are ‘smart’ fluorescence probes.

Swager and coworkers reported the first compound NIAD-4 (8, Figure 4) in 2005 [58]. Notably, in the amyloid-binding pocket, the rings of NIAD-4 are rigidly held in a close planar conformation that has better electron conjugation, thereby, fluorescence QY is improved [58,66]. However, the wavelengths of excitation/emission peaks are out of NIR region (λ_ex = 475 nm, λ_em = 612 nm, Table 2). In order to further red-shift the spectral wavelengths and increase fluorescence QYs, in 2008, the same group introduced compounds NIAD-11 and NIAD-16 (9–10, Figure 4) [59]. Compared with the structure of NIAD-4, the ring of NIAD-11 is more rigid and has an extended electron conjugation. The results show that NIAD-11 exhibits 11% fluorescence QY with Aβ aggregates, which is ten-times higher compared with NIAD-11 in saline alone. It also has an emission peak above 700 nm (Table 2) [59].

The design strategy of NIADs is a promising approach for designing novel scaffolds with consideration of molecular characteristics of both probe and binding pocket. A thin hydrophobic groove along the long axis of the filament of fibrils has been identified as the binding site of fluorochromes [67]. It is an essential strategy to utilize the constraints of the fibril-binding pocket to enhance fluorescence smart probes for Aβ in future research.

NIRF probes derivatived from natural product curcumin

Fluorescent compounds named CRANADs are the first designed NIRF probes derivated from natural product curcumin, which has high binding for plaques. Ran et al. reported a series of NIRF probes by modifying the structure of curcumin. Studies demonstrated that curcumin displayed high-affinity binding for Aβ aggregates (K_d = 0.20 nM). However, curcumin is not suitable for in vivo imaging because of its short emission wavelength, limited BBB penetration and rapid metabolism [68,69]. On the basis of the structure of curcumin, Ran and coworkers employed 2,2-difluoro-1,3,2-dioxaborines as an acceptor to bridge two N,N′-dimethyl groups on aromatic rings that serve as donor groups, which are beneficial for electron delocalization in the whole structure (Figure 5). In addition, replacing the phenolic hydroxyl groups with some donor groups, such as N,N′-dimethyl, not only enables red-shifted absorption but also improves the lipophilic of curcumin for its reasonable BBB penetration. In addition, the replacement can avoid rapid metabolism, thus increase the in vivo stability of curcumin.

Figure 5. . Design philosophy of CRANAD-2.

Introduction of a difluoroboronate group into curcumin to form a red-shifted dye, and employing 2,2-difluoro-1,3,2-dioxaborines as an acceptor and two N,N′-dimethyl groups on aromatic rings as donor groups.

CRANAD-2, the first reported curcumin-derived probe, possesses considerable good optical properties (11, Figure 5 & Table 3) [49]. Upon interacting with Aβ aggregates, CRANAD-2 underwent a range of changes, such as a 70-fold fluorescence intensity increase, a 90 nm blue shift (from 805 to 715 nm) and a large increase in QY. Meanwhile, this probe also shows a high affinity for Aβ aggregates (K_d = 38.0 nM), a reasonable log p-value (LogP = 3) and high in vivo metabolic stability (Table 3) [49]. In brief, CRANAD-2 nearly meets all requirements of an NIRF probe and can be potentially applied for drug screening. In subsequent research, based on CRANAD-2, Ran et al. reported more curcumin analogs with better optical properties. Results suggested that these derivatives were capable of detecting soluble and insoluble Aβs both in vitro and in vivo. We provided more details in the following section (see the ‘NIRF probes for binding to oligomers non-specifically’ section) of NIRF probes for imaging amyloid-β oligomers, such as CRANAD-3, CRANAD-28 and CRANAD-58.

Table 3. . Binding and optical properties of the near-IR fluorescence probes from scaffolds 4–13.

No.	Scaffold	Name	MW	λex/λem	Fold increase‡	Affinity (nM)	LogP	Φ (%)	Ref.
11	4	CRANAD-2	410.26	640/802†	70	38.0	3	0.6/40§	[49]
12	5	BAP-1	351.17	614/648¶	–	44.1	–	46.8¶	[70]
13	5	BAP-2	357.23	650/708¶	–	55	–	11.4¶	[71]
14	5	BAP-3	341.16	663/705¶	–	149	–	4.5¶	[71]
15	6	BAP-4	433.32	636/704¶	–	27	–	9.3¶	[71]
16	6	BAP-5	417.26	649/723¶	–	18	–	4.3¶	[71]
17	7	EUA-1	365.23	518/654¶	–	320	–	12#	[72]
18	7	EUA-2	483.36	597/667¶	–	230	–	7††	[72]
19	7	EUA-3	601.49	677/736¶	–	320	–	1‡‡	[72]
20	8	EUA-4	511.41	624/673¶	–	48.6	–	13‡‡	[72]
21	8	EUA-5	642.59	721/763¶	–	97	–	3‡‡	[72]
22	9	DANIR-2a	197.24	452/487†	1.5	1590	2.19	0§§	[73]
23	9	DANIR-2b	223.27	539/577†	2.9	35.8	2.79	0.66§§	[73]
24	9	DANIR-2c	249.31	597/665†	12.0	26.9	3.37	4.09§§	[73]
25	10	DANIR-3a	247.29	614/538¶¶	331	44.8	3.46	4.8/0.1§	[74]
26	10	DANIR-3b	273.33	682/615¶¶	716	8.8	4.05	29.9/0.3§	[74]
27	10	DANIR-3c	299.37	783/678¶¶	280	1.9	4.56	9.0/0.009§	[74]
28	10	DANIR-3d	325.41	828/716¶¶	5	–	5.10	3.7/0§	[74]
29	10	DANIR-3e	351.44	830/-¶¶	–	–	5.62	0.1/0§	[74]
30	11	DANIR-8a	291.35	616/550¶¶	57	–	–	10‡‡	[75]
31	11	DANIR-8b	317.38	683/619¶¶	321	38.5	3.69	22.3‡‡	[75]
32	11	DANIR-8c	343.42	798/678¶¶	629	14.5	4.30	7.1‡‡	[75]
33	11	DANIR-9a	321.37	613/573¶¶	3	–	–	7.5‡‡	[75]
34	11	DANIR-9b	347.41	685/642¶¶	17	197	3.04	15.7‡‡	[75]
35	11	DANIR-9c	373.45	787/673¶¶	194	19.9	3.73	9.3‡‡	[75]
36	12	Probe 5	227.26	400/532##	49.9	1590	2.55	–	[76]
37	12	Probe 6	255.31	440/554##	58.9	2300	3.24	–	[76]
38	13	ANCA-6	452.54	410/545##	7.7	1400	3.81	–	[77]
39	13	ANCA-14	467.56	385/530##	5.0	4600	2.79	–	[77]
40	13	ANCA-15	454.52	380/525##	5.1	13800	2.74	–	[77]
41	13	ANCA-16	497.58	430/540##	2.9	6700	2.53	–	[77]
42	13	ANCA-17	496.60	410/550##	8.4	1600	3.60	–	[77]
43	13	ANCA-18	434.53	–	–	–	–	–	[77]
44	13	ANCA-19	394.46	410/535##	6.6	1600	3.14	–	[77]

^†λex and λem were measured in PBS.

^‡Fold increase in fluorescence intensity upon binding with Aβ aggregates.

^§QYs were measured in PBS/Aβ40 aggregates respectively.

^¶Fluorescence excitation and emission and QY of compounds were determined with 10 μM of the compounds in CHCl₃.

^#In reference to rhodamine 6G in H₂O (λ_exc = 488 nm, Φ_F = 95%).

^††In reference to sulforhodamine in ethanol (λ_ex = 550 nm, Φ_F = 90%).

^‡‡In reference to crystal violet in methanol (λ_ex = 610 nm, Φ_F = 66%).

^§§QYs were determined using a calibrated integrating sphere. QYs were determined in dichloromethane.

^¶¶Determined in PBS (λ_em1) and upon binding with Aβ aggregates (λ_em2).

^##λex and λem were measured upon binding with Aβ aggregates.

λem: Maximum emission wavelength; λex: Maximum excitation wavelength; Aβ: β-amyloid; MW: Molecular weight; QY: Quantum yield.

Modification of natural products could be a promising method to search novel NIRF probes. Especially, researchers should focus on the natural products that bear therapeutic effects on AD and possess large conjugation system. These lead compounds can be easily modified to be dual functional probes that are not only useful for diagnosis but also have the capacity to treat the disease, which could be an active area of research over the next few years.

NIRF probes derivatived from boron dipyrromethene

Since first reported in 1968, boron dipyrromethene (BODIPY) and its derivatives, especially in the last decade, have been proposed for applications in many diverse areas such as molecular sensors/logic gates, photodynamic therapy, dye-sensitized solar cells and light harvesting/energy transfer cassettes [78–80]. In 2009, Suzuki and coworkers reported a new series of long-wavelength fluorescent dyes based on BODIPY with sophisticated optical properties [81], named Keio Fluors. More importantly, Keio Fluor dyes allow the easy and fine tuning of absorption/emission peaks by changing the substituents attached to the BODIPY cores, which is helpful to establish a compound library of BODIPY fluorescent dyes.

Ono et al. developed dual-modal (SPECT and NIRF) probes [¹²⁵I] BODIPY7 (Figure 6) based on the structure of BODIPY in 2011 [82]. Though the dual-modal probes have been applied successfully in the field of cancer imaging [83], such design strategy for imaging Aβ plaques has not been reported before. In theory, dual functional nuclear/fluorescent imaging probes can provide complementary information to improve diagnosis and management of AD patients, and facilitate the validation of optical imaging by standard nuclear imaging techniques. However, shorter wavelengths of absorption/emission and poor brain uptake indicate that further optimization of the structure is necessary.

Figure 6. . Near-IR fluorescence probes derivatived from boron dipyrromethene.

(A) The structure and properties of BODIPY7; (B) design strategies of new series of BODIPY derivatives.

BODIPY: Boron dipyrromethene.

To overcome the above limitations of BODIPY7, in 2012, Ono et al. introduced dimethylamino styryl group to improve not only wavelengths but also affinity for Aβ plaques (12, Figure 6) [70]. However, obviously, the excitation and emission wavelengths of BODIPY-based probe-1 (BAP-1) were still shorter than the ideal wavelengths for optical imaging in vivo (Table 3). Meanwhile, results also showed that BAP-1 had nonspecific accumulation both in Tg2576 mice and age-matched controls. Based on these findings, Ono et al. designed and synthesized novel BAPs for NIRF imaging in vivo: BAP2–5 (13–16, scaffold 5–6, Figure 6) [71]. According to the previous report of Suzuki et al., they replaced a phenyl group with a thiophenyl or furanyl group to extend the wavelengths of excitation and emission. Results demonstrated that BAPs had better optical properties than BAP-1: BAPs exhibited excitation and emission wavelengths of 636–663 nM and 704–723 nM, respectively. In addition, ex vivo fluorescent staining of brain sections of Tg2576 mice after the injection of BAP-2 showed selective binding of Aβ plaques with low nonspecific binding.

Aminophenylethenyl unit is a ‘star’ group, which appears in many probes for Aβ plaques, such as Th-T, CRANAD-2 and some PET probes. Ono et al. investigated the function of aminophenylethenyl in different position of BODIPY core [72]. The first set (17–19, scaffold 7) of the dimethylaminostyryl groups is placed at the meso position of the BODIPY core, and the second (20–21, scaffold 8) is at the 3- and 5-positions of BODIPY core. Among these probes, EUA-5 have the longest excitation/emission wavelength (721 nm/763 nm); EUA-1, EUA-2 and EUA-4 show considerably good staining patterns in the brain sections of Tg2576 mice; especially EUA-4 possesses relatively a good affinity (K_d = 48.6 ± 10 nM).

Rich chemistry and easy derivatization at every position on the core of BODIPY dyes are attractive properties for being new NIRF probes scaffolds. Future derivatization strategy should focus on balancing the polarity of compounds and obtaining desirable emission properties.

NIRF probes derivatived from the classic donor–acceptor architecture

The donor–acceptor architecture bridged by a conjugated π-electron chain is an important class of organic nonlinear optical chromophores, and this architecture has been selected as the backbone structure in the NIRF probes of donor−acceptor NIR probes (DANIRs), which were designed by Cui et al. (Figure 7). DANIR probes possess three kinds of scaffold that are likely to be the simplest Aβ-binding structures. Compared with those complicated compounds, DANIRs provided some useful information on the interaction between small molecules and Aβ fibers, and new insight of design strategy of AD probes.

Figure 7. . Near-IR fluorescence probes derivatived from the classic donor–acceptor architecture.

(A) Structural modes of DANIR probes and three reported scaffolds by Cui and coworkers. (B) Representative images of transgenosis (APPswe/PSEN1) mouse (top row) and wild-type control mice (bottom row) at different time points before and after intravenous injection of DANIR-3C/8C, which indicates 8c possesses comparatively good brain kinetics [74,75].

Possessing only one benzene ring is the most obvious characteristic of the scaffold 9 (22–24, Figure 7), which significantly reduces the lipophilicity and MW compared with CR and Th-T [73]. While the high affinity to Aβ aggregates of DANIR 2b–c (22–24) indicated that these small and planar structures were likely bound to the clefts on the surface of the Aβ fibrils. In addition, DANIRs showed high brain penetration in vivo and fast washout kinetics from normal brain regions. A direct comparison of the brain kinetics between 2c and the BAP-1 showed that probe 2c (brain_{2 min}/brain_{30 min}: 5.66) cleared much faster than BAP-1 (brain_2min/brain_30min: 1.82). Compared with CRANAD-2 (K_d = 38.7 nM), the probes of scaffold 9 exhibited outstanding affinities for Aβ aggregates (2b, K_d = 35.8 nM; 2c, K_d = 26.9 nM, Table 3). However, the short π-conjugation system led to short excitation and emission wavelengths, and thus the compounds of scaffold 9 were not suitable for in vivo imaging.

In order to shift the emission wavelength to the NIR range and improve the values of QY, Cui and coworkers modified the structure of scaffold 9 by replacing the benzene ring with a naphthalene ring [74,75]. Results showed that the emission wavelengths of probe 3b–e (25–29, scaffold 10, Figure 7) red-shifted to the NIR range (>680 nm in PBS) with large Stokes shifts (>120 nm). In addition, as reported, for aromatic hydrocarbons such as benzene, naphthalene, anthracene and pyrene, a remarkable increase in QY was observed when the size increased from 1 to 4 rings [84]. It was reasonable to speculate that the value of QY could be improved by replacing the benzene ring with a naphthalene ring. Meanwhile, multiple torsions along the single bonds of the polyenic chains can remarkably affect the global geometry of the molecules in a solution [85]. These effects also have a noticeable impact on the QYs of the probes and the interaction between probes and Aβ aggregates. With increasing polyenic chain length, the values of QY had a remarkable decrease from 3b to 3e (Table 3).

By incorporating hydroxyethyl groups into the electron donor moiety of scaffold 10, the compounds of scaffold 11 (30–35, Figure 7A) had an improvement of hydrophilicity [75]. Compared with 3c (brain_2min/brain_60min = 3.3), 8b and 8c possess comparatively good brain kinetics (8b, brain_2min/brain_60min = 10.8; 8c brain_2min/brain_60min = 5.6). However, this kind of improvement did not fit for all cases. Meanwhile, compared with DANIR3b-c, the affinity for Aβ aggregates dropped to a certain extent (such as 8c, K_d = 14.5 nM; 3c, K_d = 1.9 nM, Table 3). In order to explain this phenomenon, the authors speculated that the larger substitutions at the N position forced the molecules to prefer geometries with greater degrees of nonplanarity that were inappropriate to fit the binding pocket. In addition, in vivo NIR fluorescent imaging revealed that 8c could efficiently distinguish between AD transgenic model mice and normal controls (b, Figure 7).

Due to better fitting to the binding channel with lower binding energies, small molecules that possess larger conjugated systems have more potential to bind to Aβ aggregates and plaques. This strategy has been successfully applied for the design of NIADs. However, the overlong conjugated double-bond bridge might play bad roles in binding of targets and QYs of the probes. Therefore, more attentions should be paid to the balance between length of π-conjunction system and properties of probes to make the molecules fill in the grooves on the fibril surface [74].

NIRF probes based on the other scaffolds

Besides the scaffolds described above, other novel molecular scaffolds can serve as smart Aβ probes as well. Though poor optical properties of these fluorescent dyes may prevent their applications from in vivo imaging, the design strategy could shed light on the future research.

NIRF probes derivatived from chalcone

Studies show that flavones dose-dependently inhibit the formation of Aβ aggregates, as well as destabilize preformed Aβ aggregates, indicating that these molecules can directly interact with Aβ aggregates [86]. In 2007, Ono and coworkers reported a series of chalcone derivatives as PET/SPECT probes for in vivo imaging Aβ plaques. These chalcone-mimic compounds showed diversified binding affinities for Aβ aggregate, which were varied from 3 to 105 nM [87]. Inspired by the work of Ono, Jung et al. designed a series of chalcone derivatives served as NIRF probes by modifying chalcone with a focus on the aromatic furan ring to improve characteristics, such as Aβ plaque affinity and fluorescent properties (36–37, scaffold 12, Figure 8). Though these probes were able to specifically stain the Aβ plaques in a brain section from a transgenic AD model mouse, especially probe 5 exhibited an approximately 50-fold increase in emission intensity after mixing with Aβ42 aggregates, low micromolar affinity (K_d = 1.59 μM) and short excitation/emission wavelength (400 nm/532 nm) prevent their application from in vivo imaging (Table 3) [76].

Figure 8. . Near-IR fluorescence probes derivatived from other scaffolds.

(A) Chemical structures of chalcone derivative; (B) structures of amino naphthalenyl-2-cyano-acrylate-based β-amyloid-binding probes.

NIRF probes derivatived from amino naphthalenyl-2-cyano-acrylate

In 2011, Chang et al. designed a new family of fluorescent probes based on the amino naphthalenyl-2-cyano-acrylate motif (38–44, scaffold 13, Figure 8). Through modifications of the hydrophilic group and nitrogen donor group, they validated donor part of the molecule, which was likely to bind the pocket of the aggregated protein [77,88]. However, in view of the low affinity and selectivity for Aβ plaque, the properties of this series of fluorescent probes were not so well fit for in vivo imaging.

Compounds for imaging Aβ oligomers

With evidence that the extent of deposited amyloid poorly correlated with cognitive impairment, researchers have shifted their focus to soluble forms of Aβ. Data indicate that small soluble oligomers of Aβ inhibit hippocampal long-term potentiation in vitro and in vivo, and induce synapse degeneration in the brain of AD patients [31,89]. These results support the hypothesis that diffusible oligomers of Aβ initiate a synaptic dysfunction, which is likely an early event in presymptomatic of AD. Compared with Aβ plaques, oligomers may be an earlier and more precise biomarker for early AD diagnosis. However, due to heterogeneous and transient nature of Aβ oligomers, it is remarkably difficult to detect these species specifically. Thus, reports related to development of NIRF probes for binding to Aβ oligomers are very rare, nonetheless a summary of these rare cases are highly necessary to provide useful information for further probe development.

NIRF probes for binding to oligomers nonspecifically

With the modification of curcumin, Ran and coworkers first designed and synthesized a family of NIRF probes for binding to various Aβ species [90]. Compared with the previously reported NIRF probes, these compounds were capable of binding soluble Aβ species, which were the likely biomarker in the early stage of AD [89,91,92].

Aβ_13–20 fragment (HHQKLVFF, as shown in Figure 9) possesses hydrophilic/hydrophobic regions and appropriate structural stereo-hindrance compatibility. As mentioned, CRANAD-2 is the first reported curcumin-based NIRF imaging probe, but it lacks the capability of detecting soluble Aβ species. Considering the hydrophobic property of CRANAD-2, its symmetric structure does not match with the hydrophobic and hydrophilic properties of Aβ_13–20 fragment [90]. In order to enhance interaction with the hydrophobic LVFF segment (Aβ_{17-2 0}, Leu17, Val18, Phe19, Phe20), Ran et al. attempted to cut the CRANAD-2 in half and obtained the compound CRANAD-54. As expected, CRANAD-54 showed significant fluorescence changes with KLVFF and much stronger fluorescence intensity increase with monomeric Aβ₄₀ than that of CRANAD-2. However, CRANAD-54 is not suitable for in vivo imaging, due to its short excitation and emission wavelengths.

Figure 9. . The schematic of binding site: Aβ_13–20 fragment (the model of Aβ_1–42 hexamer from x-ray [4NTR]).

Aβ: β-amyloid.

Based on the structure of CRANAD-54, CRANAD-58 (47, Figure 10) was designed to have a longer π-conjugation system, in which its pyridyl moiety was conjugated to match the hydrophilic HHQK segment [90]. CRANAD-58 displays excellent fluorescence properties (λ_ex = 630 nm, λ_em = 750 nm) and 91.9- and 113.6-fold fluorescence intensity increases at 672 nm for Aβ₄₀ and Aβ₄₂ monomers, respectively. However, its binding affinity declines to a certain degree compared with CRANAD-2 (Aβ40: K_d = 105.8 nM, Aβ42: K_d = 45.8 nM). In an attempt to enhance the interaction with Aβs, Ran and coworkers designed CRANAD-3 by replacing the phenyl rings of CRANAD-2 with pyridyl to introduce potential hydrogen bonds [93]. As the experimental results suggested, CRANAD-3 (46, Figure 10) exhibited strong binding with Aβ_40/42 monomers, dimers and oligomers (K_d = 24 ± 5.7 nM, 23 ± 1.6 nM, 16 ± 6.7 nM and 27 ± 15.8 nM, respectively, Table 4) [93]. Both CRANAD-3 and CRANAD-58 can differentiate transgenic and wild-type mice as young as 4-months old, the age that assembles soluble Aβ in brain. Meanwhile, its fluorescence properties nearly meet all the requirements for an NIRF probe for the detection of Aβs both in vitro and in vivo.

Figure 10. . A new series of curcumin analogs for binding to β-amyloid oligomers nonspecifically.

Table 4. . Binding and optical properties of the near-IR fluorescence probes for binding β-amyloid oligomers from scaffold 12–16.

No.	Scaffold	Name	M.W.	λex/λem†	Fold increase‡	Affinity§ (nM)	Affinity¶ (nM)	Log P	Φ (%)	Ref.
45	14	CRANAD-28	484.30	498/578	–	52.4	68.8 159.7 162.9 85.7	–	32#	[94]
46	15	CRANAD-3	468.35	605/730	12.3 39.5 16.3 16.1	–	24 23 16 27	2.50	–	[93]
47	15	CRANAD-58	439.31	630/750	91.9 113.6	–	105.8 45.8	1.94	–	[90]
48	16	BODIPY-5	481.35	520/550††	25‡‡	–	–	–	3††	[95]
49	16	BODIPY-6	537.45	529/540††	35‡‡	–	–	–	1††	[95]
50	16	Aza-BODIPY	497.34	-/670	16‡‡	–	–	–	–	[96]
51	–	BD-Oligo	571.64	580/640§§	∼15‡‡	–	480	–	8.7§§	[97]

^†λex and λem were measured in PBS.

^‡Fold increase in fluorescence intensity upon binding to soluble Aβs (Aβ₄₀ monomer/Aβ₄₂ monomer/Aβ₄₂ dimers/Aβ₄₂ oligomers) for CRANAD-3, and fold increase in fluorescence intensity upon binding to Aβ₄₀ monomer and Aβ42 monomer for CRANAD-58.

^§Binding constant to Aβ aggregates.

^¶Binding constant to soluble Aβs (Aβ₄₀ monomer, Aβ42 monomer, dimers and oligomers in sequence for) CRANAD-3/28/58.

^#QYs were measured in PBS.

^††λex and λem, and QY were measured in 10 mM TRIS-NH₄OH, pH 8.7.

^‡‡Fluorescence enhancement in the presence of ordered Aβ1–42 oligomers.

^§§λex and λem, and QY were measured in DMSO.

λem: Maximum emission wavelength; λex: Maximum excitation wavelength; Aβ: β-amyloid; BODIPY: Boron dipyrromethene; QY: Quantum yield.

However, these curcumin analogs for detecting soluble and insoluble Aβs both in vitro and in vivo always have low QY. To overcome the low QY limitation of these probes, Ran et al. replaced the phenyl rings with pyrazoles to increase the brightness [94]. As expected, CRANAD-28 (45, Figure 10) displayed high QY in PBS and ethanol. Although CRANAD-28 is an excellent two-photon imaging probe for Aβ plaques and cerebral amyloid angiopathies (CAAs), its short emission wavelengths (578 nm) prevent it from in vivo NIRF imaging.

Ran and coworkers have successfully designed a series of curcumin analogs for detecting soluble and insoluble Aβs both in vitro and in vivo. Though these compounds exhibited low selectivity between Aβ subspecies, some have showed the potential for monitoring Aβ species at a presymptomatic stage of AD. In addition, the structure–activity relationship studies of these curcumin analogs are very important to guide further work about designing more selectivity NIRF probes.

Based on previous research work, our group has designed a series of NIRF probes with the curcumin scaffold to selectively detecting soluble Aβs. This development is currently underway.

NIRF probes for binding to Aβ oligomers specificity

Due to Aβ oligomers’ heterogeneous and transient nature, it is considerably difficult to selectively detect these species. As stated before, BODIPY dyes could be used as NIRF probes for amyloid aggregates when they were attached to a pharmacophore [70–72,79–82,95]. And in 2009, Dzyuba and coworkers first described novel triazole-containing BODIPY dyes (48–50, scaffold 16, Figure 11) to distinguish the unordered and ordered conformations of soluble Aβ_1–42 oligomers [95]. The investigation of triazole-containing BODIPY dyes showed that the dyes provided up to an 8-fold and 35-fold fluorescence increase in the presence of the unordered and ordered, β-sheet-rich conformations of soluble Aβ_1–42 oligomers, respectively. However, the emission spectra of these compounds are too short to meet the requirements of classic NIRF probes (Table 4). In 2013, the same group reported the compound Aza-BODIPY (50, Figure 11), which has the wavelengths of excitation/emission maxima above 600 nm, owning to larger conjugated systems [96]. Results showed that the fluorescence intensity of the dye increased 6-fold in the presence of unordered soluble oligomers and 16-fold when the ordered form was added. However, these BODIPY analogs lack capability of discriminating oligomers from fibrils.

Figure 11. . Chemical structures of triazole-containing boron dipyrromethene dyes and Aza-boron dipyrromethene that were reported by Dzyuba and coworkers.

In 2015, BoDipy-Oligomer (BD-Oligo) (51, Figure 12A) was reported by Chang and coworkers as the first oligomer-specific sensor [97]. BD-Oligo is picked out from high-throughput screening of 3500 fluorescent candidates. With BD-Oligo, the highest fluorescence enhancement is observed upon incubation with Aβ oligomers. During the path of fibril formation, the fluorescent signal of BD-Oligo increased as monomers aggregate into oligomers but decreases as more Aβ assemble into fibrils, which indicated that this dye had oligomer-sensing ability (Figure 12B). Though BD-Oligo successfully penetrates BBB and shows Aβ oligomers detection capabilities in the brains of the AD transgenic mice model without toxicity, the low binding affinity (K_d = 0.48 μM, Table 4) and short wavelength of excitation (λ_ex = 530 nm, Table 4) definitely hinder the application of BD-Oligo in vivo NIRF imaging.

Figure 12. . The properities of compound BD-Oligo.

Chemical structure of BD-Oligo, and absorbance maximum, emission maximum and quantum yield of BD-Oligo, measured in DMSO; emission spectra indicate that BD-Oligo shows the different fluorescence enhancement when incubated with monomers, oligomers and fibrils of Aβ, and which proves that BD-Oligo is oligomer-specific sensor.

Aβ: β-amyloid.

Data taken from [97].

The first attempt of designing of oligomer-specific NIRF probe seems unsuccessful, but this tedious approach provides a model for the interactions of BODIPY with Aβ oligomers and a good starting point for further NIRF probe development applicable.

Conclusion

Noninvasive molecular imaging of Aβ plaque plays a key role in the clinical assessment of patients with suspected AD. Compared with established imaging modalities, such as positron emission tomography (PET) and MRI, near-IR fluorescence imaging has many advantages, such as low cost, high throughput, and easy operation. In this review, we focused on recently reported NIRF probes that showed excellent binding to plaques and oligomers, and disscussed the advantages and disadvantages of these probes as NIRF tracers. Substantial research work has been dedicated to developing NIRF probes binding to Aβ plaques, but research on probes that can bind to Aβ oligomers, the most toxic Aβs, are very limited. We hope that this review will promote more research to develop imaging probes that can bind to soluble Aβs. With such probes, early diagnosis of AD could be feasible.

Future perspective

The exploitation of more NIRF probes that are selectively targeting Aβ oligomers may be a priority for future research. Studies have suggested that soluble oligomers play key roles in the early phase of AD before the formation of the plaques. Thus, oligomers are likely the earlier and more precise biomarkers for AD. Meanwhile these oligomers-specific probes that are capable of monitoring Aβ species at the presymptomatic stage of AD are beneficial to the development of anti-Aβ drugs and pathological studies. To date, BD-Oligo is an ‘orphan’ probe for binding oligomers, which is obviously inadequate. Except with the help of high-throughput screening from fluorescent libraries, the new structures can be derived from existing scaffolds, such as Th-T and CR, which are capable of binding to oligomers with high affinities but no specificity over insoluble Aβs [98]. In addition, the modification of established NIRF probes is faster and more efficient way to obtain new targeted probes.

Bifunctional NIRF probes may be another active area of research over the next few years. Natural products with potential therapy for AD have provided several new scaffolds for probes in the last decades. Data indicate that mangosteen, curcumin and flavonoids are capable of inhibiting the formation of Aβ aggregates and destabilizing preformed Aβ aggregates, which suggest that the NIRF probes derived from these natural products may possess therapeutic potential. As mentioned above, curcumin analog CRANAD-28 is not only useful for two-photon imaging but also has the capacity to attenuate crosslinking of Aβ42.

Converting the NIRF probes to PET tracers may be a promising strategy to promote the application of NIRF probes in a clinical setting. Incorporation of isotopes into NIRF molecules will facilitate the use of the probe for PET imaging. In theory, dual functional nuclear/fluorescent imaging probes can provide complementary information, and facilitate the validation of optical imaging by standard nuclear imaging techniques.

Last, rapid advances in fluorescent imaging technology are likely to provide a huge leap for NIRF probe development. Although NIRF imaging has been only used in brain imaging on small animals, and it is still difficult for imaging of Aβs inside the human brain. Fluorescence molecular tomography, which is potentially capable of detecting signals at depths of 7–14 cm, has the potential capacity to translate NIRF imaging into clinical applications in future [99,100].

Executive summary.

Background

• Currently, no treatment is available to stop/slow the disease's progression, and most promising strategy for treating Alzheimer's disease (AD) might be to the early intervention is based on early diagnosis at the early stage.

• Based on amyloid cascade hypothesis and its modification version, both β-amyloid (Aβ) deposition and soluble oligomeric species may be good diagnostic biomarkers for AD.

The development of molecular imaging technologies

• Noninvasive molecular imaging of Aβ plaque plays a key role in the clinical assessment of patients with suspected AD, and MRI and positron emission tomography are established imaging techniques for diagnosis of AD. However, the molecular imaging technologies that have been currently applied to imaging Aβs can not completely fulfill the needs of clinical diagnosis.

• Compared with positron emission tomography and single-photon emission computed tomography imaging modalities, near-IR fluorescence imaging has many advantages, and it may have the potential for clinical use.

• Substantial research work has been dedicated to developing near-IR fluorescence probes binding to Aβ plaques, and CRANAD-Xs and boron dipyrromethenes are the most representative probes among the reported structures.

• NIRF probes for selectively binding to Aβ oligomers are very limited, and more investigations for developing oligomer-specific probes are highly desirable.