Rod-like amphiphilic AIE-active NIR probe enables super-early precise in-vivo detection of Aβ brils/plaques

Precise and early detection of Aβ brils/plaques is pivotal to the diagnosis and treatment of Alzheimer's disease (AD), a serious disease threatening human health. Optical imaging stands out to be a promising technique for such task. However, restricted by poor blood-brain barrier penetrability, short-wavelength excitation and emission, and aggregation-caused quenching effect, the clinically used gold-standard probes are usually powerless in early in-vivo diagnosis of AD. To address these issues, we put forward an “all-in-one” design principle and develop a simple rod-like amphiphilic NIR AIE probe to demonstrate its feasibility. In-vitro, ex-vivo, and in-vivo experiments with different strains of mice indicates that AIE-CNPy-AD holds the universality to Aβ brils/plaques identication. Noteworthily, AIE-CNPy-AD is even able to precisely trace the small and sparsely-distributed Aβ brils/plaques in AD model mice as young as 4-month-old APP/PS1 mice, the youngest having Aβ deposits, suggesting the probe might be an ideal alternative for early AD diagnosis. The cell viability was calculated using GraphPad Prism 7.0 software. The cytotoxicity assay procedures of mouse brain neuroblastoma cells (Neuro-2a), mouse breast cancer cells (4T1), and human breast cancer cells (MCF-7) are the same as that of SH-SY5Y, except that the number of the other three cells inoculated in 96-well plates was 10,000.


Introduction
Alzheimer's disease (AD) is a degenerative disease of the nervous system, which will lead to memory loss, behavior disorder, cognitive decline, and eventually death [1][2][3] . According to the statistic, there are about 50 million patients with Alzheimer's disease all over the world. It is predicted that the number of AD patients will increase to 150 million by 2050, exerting serious social and economic burden to countries worldwide 4,5 . Currently, clinical diagnosis of AD patients is mainly through the combination of the inquiring of the patients' genetic history, neuropsychiatric test, neuropathological diagnosis, and neuroimaging diagnosis. Only after death of AD patients, can nal diagnosis be con rmed by brain tissue examination 6,7 . β-Amyloid (Aβ) hypothesis has shown that the extracellular accumulation of β-amyloid peptides forming brotic plaques is one of the neuropathological hallmarks of AD 8-13 , as such early detection of Aβ brils and plaques in vivo with low damage plays extremely important role in the diagnosis and subsequent treatment and prevention of AD 9,14-20 .
By far, a large number of imaging techniques, such as magnetic resonance imaging (MRI), positron emission tomography (PET), single photon emission computed tomography (SPECT), have been utilized to clinically diagnose AD 21 . Nevertheless, MRI can only image large plaques because of its limited sensitivity; PET needs radioisotopes which are hard to be obtained and probably impose radioactive exposure danger to patients; and SPECT has relatively higher background noise and the tracers used in this technique usually have poor blood-brain barrier (BBB) penetrating ability 1,6,15,22,23 . Drawbacks of these techniques hamper their wide application in early diagnosis of AD 11 . Compared with the above techniques, uorescence imaging technique enjoys real-time and in-situ monitoring ability, high sensitivity, low biological toxicity, non-invasiveness, superior spatiotemporal resolution, low cost, and technical simplicity, making it promising in the imaging of Aβ brils and diagnosis of AD 1,7,11,23,24 .
ThT, ThS, and indocyanine green (ICG) are the most commonly used uorescent dyes to histologically stain Aβ brils and plaques 25 . However, the high water-solubility of these dyes limits their ability of penetrating the BBB 26, 27 , rendering them merely effective in in-vitro and ex vivo imaging of Aβ brils/plaques. Moreover, ThT is hardly able to distinguish Aβ peptides of different aggregation levels 28 .
The short-wavelength excitations and emissions of probes like ThT and ThS prevent them from being used for in-vivo imaging, because of the unsatisfactory deep-tissue penetration ability and non-negligible phototoxicity to organisms. In addition, most of these uorophores suffer from aggregation-caused quenching (ACQ) effect that results in self-quenching of FL signal after binding to Aβ species and the reduced detection sensitivity and imaging resolution 29 . These objective factors such as the poor speci city, less satisfactory sensitivity, small Stokes shift, relatively low reliability make these probes di cult to ensure their effectiveness in in-vivo imaging 7,8,30 . Sensitive and reliable probes with high ability to penetrate BBB and deep tissues which can be employed to detect Aβ brils and image Aβ plaques in vivo is urgently desirable 31 .
By analyzing the structure and performance of the small-molecule uorescent probes reported to detect Aβ brils and plaques, it can be found that the ones having rod-like geometric con gurations and donoracceptor (D-A) electronic structures usually perform relatively well ( Fig. 1A) 1,2,14,26,32 . On the one hand, the rod-like structured probes have a certain similarity with Aβ brils in morphology, which is conducive to the binding of probes and Aβ brils and subsequently bene ts the speci city. On the other hand, the probes with D-A effect are sensitive to the environmental hydrophobicity, with the emissions intensi ed when bound to hydrophobic areas of amyloids abundant in β-sheet structures [33][34][35] . Though these two essential structural features can guarantee the basic performance of the probes, other de ciencies still seriously affect the detection and imaging performance of these probes. Most of the reported probes are ACQactive, resulting in high background and low SNR. Moreover, simple D-A structure in probes (e.g. ThT) cannot ensure red or NIR emission, which can neither avoid the interference from auto-uorescence of organisms nor assure the penetration ability of deep tissue.
Based on the above analysis, we sort out four main criteria which should be satis ed to realize highperformance uorescent probes for detecting and imaging Aβ brils and plaques in vivo at an early stage ( Fig. 1B): (1) geometric con gurations which match the β-sheet structure; (2) balanced hydrophilicity and hydrophobicity which guarantees the BBB crossing ability and high SNR; (3) strong D-A effect ensuring uorescent response, long-wavelength excitation and emission, and large Stokes shift; (4) anti-ACQ effect that enhances the uorescent response, SNR, and imaging resolution. In accordance with these criteria, we propose a systematic "all-in-one" strategy for rational design of high-performance in-vivo imaging contrast agents for precise detection of Aβ brils/plaques. In other words, rod-like geometric con gurations, hydrophilic unit-decorated hydrophobic skeleton, large D-π-A electronic structure, and 3D exible conformation are ingeniously integrated in one molecule to meet the above four criteria (Fig. 1B)..
As a proof of concept, AIE-CNPy-AD is designed following the "all-in-one" strategy. In this molecule, electron-donating dimethylamino, electron-accepting acetonitrile and pyridyl group are linked together via single bonds and bridged by benzyl groups and a C=C double bond, affording a large exible 3D rod-like con gured D-π-A framework. Bene ted from such a framework, strong Aβ bril/plaque-binding capability, e cient aggregation-induced emission (AIE) effect 36 and bright red/NIR uorescence are achieved. Furthermore, the hydrophilic propanesulfonate group is attached to the hydrophobic molecular backbone to generate the zwitterionic and amphiphilic AIE-CNPy-AD, endowing AIE-CNPy-AD with low background, high SNR and good BBB penetration ability. Since as compared with traditional ACQ probes, AIE probes often have the advantages of good photo-stability, high SNR, and resistance to photo-bleaching, the imaging performance of the uorescent probes is greatly improved [37][38][39][40] . As a result, the rod-like amphiphilic NIR-emissive AIE probe AIE-CNPy-AD possesses excellent speci city to Aβ brils/plaques, superb deep-tissue penetrating ability, good resistance to auto-uorescence interference from organisms, improved photo-stability, superior BBB penetrating capability, and high-resolution and high-contrast imaging ability.
With the aid of the above merits, the elaborated probe AIE-CNPy-AD is able to realize detection of Aβ brils in vitro with high SNR and in-situ mapping of Aβ plaques in vivo with high sensitivity, and high delity and contrast. More importantly, the precise in-situ and in-vivo mapping ability of Aβ plaques is not limited by mouse strains. It is worth mentioning that even small and sparsely distributed Aβ plaques in the brains of AD transgenic mice APP/PS1 as young as 4-months old could be visualized by our probe. It is reported that the Aβ plaques would not appear when the APP/PS1 mice are younger than 4-months old. Moreover, the increase and enlargement of Aβ plaques as well as the progression of AD as the mice grow could be clearly revealed by the present probe. In other words, AIE-CNPy-AD is very promising in early diagnosis and highly reliable progression monitoring of AD (Fig. 1C). To the best of our knowledge, this is the rst work of harnessing the "all-in-one" strategy to rationally design AIE-active NIR imaging contrast agent for light-up and in-situ tracing of Aβ plaques in mice of different strains and ages. What's more, the probe maps the Aβ plaques at the earliest stage among all the reported uorescent probes.

Results
Rational design and facile synthesis of AIE-CNPy-AD. We make full use of the "all-in-one" strategy to pursue high-performance probes for early in-situ tracing of Aβ brils in vivo. As introduced above, extended D-π-A electronic architecture is built by utilizing the dimethylamino as electron-donor, the acetonitrile and pyridyl group as electron-acceptor, single bonds as linkers, and benzyl groups and C=C double bond as π-bridges, respectively. Red or even NIR emission and relatively long-wavelength excitation is supposed to result from this D-π-A structured backbone. Such a rod-like geometric con guration is envisioned to bestow the probe with strong binding a nity to Aβ bril/plaque. Moreover, according to the principle of restriction of intramolecular motions (RIM), the multiple rotors in the skeleton would e ciently consume the excited-state energy and lead to weak or even no emission in the unconstrained state, while the distorted 3D conformation would prevent π-π stacking and self-quenching in the constrained state. It means such an AIE-active probe would give light-up response to the target species with satisfactory SNR under suitable conditions. To further reinforce the SNR, hydrophilic propanesulfonate group is incorporated to the hydrophobic skeleton to for one thing enhance the solubility of the probe in aqueous media and thus reduce the background/noise signal, and for another to strengthen the binding between the probe and Aβ brils/plaques via multiple electrostatic interactions.
Moreover, the considerable lipophilicity of the AIE-CNPy-AD is anticipated to be su cient to ensure desirable BBB penetrability. The elaborately designed rod-like amphiphilic NIR-emissive zwitterionic AIE probe AIE-CNPy-AD was conveniently synthesized according to the synthetic route shown in Fig. S1 in the Electronic Supporting Information (ESI) with commercially available cheap raw materials. AIE-CNPy-AD was fully characterized with the help of 1 H (Fig. S2), 13 C NMR (Fig. S3) and high-resolution mass spectrometry (HRMS, Fig. S4).
Excellent photophysical properties facilitating the detection of Aβ brils/plaques. AIE-CNPy-AD showed an absorption maximum at 455 nm and an emission maximum at 720 nm (10 -4 M) in DMSO solution ( Fig. 2A), with the Stokes shift of as large as 265 nm. As AIE-CNPy-AD is soluble in highly polar solvents but aggregates in less polar solvents, DMSO was chosen as a good solvent, and THF was selected as a poor solvent to evaluate the AIE behaviors of AIE-CNPy-AD ( Fig. 2 and Fig. S5). As anticipated, with the addition of THF, uorescence enhancement was clearly observed, with the emission peak blue-shifted from 720 nm (DMSO solution) to 675 nm (DMSO/THF = 1/99, v/v), manifesting the synergistic effect of AIE and intramolecular charge transfer (ICT) (Fig. 2B). The emission intensity continuously and slowly increased when the THF fraction (f THF ) was no more than 60 vol%, while boosted sharply once f THF reached 70 vol% (Fig. 2C). Remarkably, AIE-CNPy-AD was hardly affected by the pH value varying from 3 to 10 which covers the normal pH range of human body ( Fig. S6 and Fig. 2D), suggesting the good pH stability of the uorescence properties of our probe. Taken together, all these results undoubtedly demonstrated the remarkable AIE feature, strong ICT effect, and the stable NIR uorescence in the aggregated state of AIE-CNPy-AD, which are conducive to the imaging of Aβ deposits.
Superb speci city and high a nity of AIE-CNPy-AD to Aβ brils. The binding a nity of AIE-CNPy-AD to hen egg white lysozyme (HEWL), a common model protein for amyloid studies, was evaluated rst. As shown in Fig. 3A, dramatic uorescence enhancement at 620 nm was clearly observed with the increasing concentration of brillar HEWL in the PBS solution. In sharp contrast, AIE-CNPy-AD displayed much less marked uorescence response to native HEWL at the same concentration level (Fig. S7A). As compared to the emission in DMSO solution, the probe bound to HEWL showed an emission peak blueshifted from 720 to 620 nm, which might be attributed to the ICT effect and less polar microenvironment of the protein pockets. To our satisfaction, AIE-CNPy-AD has much higher ability to distinguish brillar HEWL from native ones, as ThT had the same response to brillar and native HEWL (Fig.s 3B, 3C, and We then studied whether this AIE-active NIR probe could have a speci c uorescence response to Aβ brils. Aβ 1-42 peptide was fully incubated to afford Aβ 1-42 brils with expected brous or lamentous structure (Fig. 3D). When the Aβ 1-42 brils were added, emission intensity of AIE-CNPy-AD increased consistently with the emission peak blue-shifted to 620 nm like the situation of binding to HEWL (Fig. 3E). Obviously, the background from AIE-CNPy-AD is far lower than that of ThT, which is merely 1/35 times that of ThT. Whereas, the signal of AIE-CNPy-AD upon interaction with Aβ 1-42 brils is much higher than that of ThT under parallel conditions, which is about 1.4 times that of ThT (Fig. S8). In consequence, the SNR of AIE-CNPy-AD reached 60, which is 10 times to that of ThT (SNR ThT = 6; Fig. 3F). The signi cantly enhanced SNR of AIE-CNPy-AD achieved by the minimization of background and ampli cation of signal could be interpreted as follows: (i) on one hand, the right hydrophilicity makes the AIE-CNPy-AD well dispersed in aqueous solution and the vigorous intramolecular motions e ciently exhaust the excited state, resulting in minimal emission; (ii) on the other hand, the strong binding of AIE-CNPy-AD to Aβ  brils greatly hampered the intramolecular motions and activated the radiative decay channels, maximizing the uorescence signal as a consequence of AIE effect. Ultrasensitive detection of Aβ  brils could be expected.
In addition to the ultra-high SNR, AIE-CNPy-AD also has high speci city to Aβ 1-42 brils. A large variety of biological species including carbohydrates, amino acids, peptides, and other proteins were employed to assess speci city of AIE-CNPy-AD to Aβ 1-42 brils (Fig.3G, Fig. S9). It is evident that AIE-CNPy-AD not only hardly has response to interfering small molecular species, but also has low response to potentially competitive peptides and enzymes with large molecular weight, especially Aβ 1-42 monomer ( Fig. 3G and Apart from speci city, binding a nity of the probe to analyte is also a vital parameter that ensures accurate tracing of the Aβ 1-42 brils. Displacement assay of AIE-CNPy-AD against ThT-bound Aβ 1-42 brils ( Fig. S10 and Fig. 3H) was then carried out to investigate the binding a nity. Fluorescence intensity of the pre-prepared ThT/Aβ 1-42 brils complex was rstly recorded with excitation at 420 nm.
AIE-CNPy-AD solution was subsequently added stepwise into the ThT/Aβ 1-42 complex, and the emission intensities of these two probes were measured under excitation at their corresponding maximum absorption wavelengths. It is observed that the emission intensity of ThT at 482 nm decreased continuously with increasing concentration of AIE-CNPy-AD; and in the meantime, uorescence of AIE-CNPy-AD peaked at 620 nm emerged and was enhanced accordingly. Remarkably, it is indicated that AIE-CNPy-AD displaced ThT from ThT/Aβ brils complex to generate the more strongly bound AIE-CNPy-AD/Aβ brils complex in the solution. Besides, dissociation constant (K d ) 41 of AIE-CNPy-AD was calculated to be 185 nM (Fig. S11), considerably smaller than that of ThT (890 nM) 42 . Su cient evidences proved that AIE-CNPy-AD has much higher binding a nity to Aβ 1-42 brils than ThT.
High-contrast and high-resolution in-vitro fluorescent staining of para n slices of mice brains. It is con rmed that AIE-CNPy-AD, with high speci city, binding a nity and SNR, exhibits excellent performance on precise detection of Aβ 1-42 brils in solution. To explore the ability of AIE-CNPy-AD to label Aβ plaques in brain tissues, in-vitro uorescent staining of para n mice brain slices resected from 5*FAD transgenic mice, APP/PS1 transgenic mice and age-matched wild-type mice was conducted. False signals originated from the binding of AIE-CNPy-AD or antibody with proteins or interfering species in brain cells could be identi ed through the localization of nuclei with Hoechst 33342. Notably, as displayed in Fig. 4 and Fig. S12, no matter in 2.5-month-old 5*FAD transgenic mice or in 6-month-old APP/PS1 transgenic mice, Aβ plaques were unambiguously visualized with bright red uorescence. Speci c labeling of Aβ plaques with high contrast and high resolution in these mice indicates that AIE-CNPy-AD is universal to label Aβ plaques in different strains of mice. Moreover, it can be easily seen from the CLSM images, the Aβ plaques in 2.5-month-old 5*FAD transgenic mice ( Fig. 4A and Fig. S12A-S12C) are larger and more densely distributed than that in 6-month-old APP/PS1 transgenic mice ( Fig. 4F and Fig. S12D-S12F). The experimental results agreed well with situation of early plaque formation at the time point in these two strains of mice 43,44 , implying the reliability of AIE-CNPy-AD in uorescent staining of Aβ plaques in slices of mice brain. To our delight, there was no observable intracellular uorescence signal, which implied that our probe is preferentially bound to the Aβ plaques generally forming extracellularly, eliminating the false signal from intracellular species. Simultaneously, the signals in the red channel agree fairly well with those in the green channel, manifesting that AIE-CNPy-AD has very high speci city comparable to that of antibody, to Aβ plaques ( Fig. 4B-4E, 4G-4J). In contrast with transgenic mice, no plaques were found in the age-matched wild-type mice no matter observing from red or green channel (Fig. S13), which not only veri ed the high speci city of AIE-CNPy-AD to Aβ plaques for a second time, but also suggested the fairly high delity of the designed probe in real brain tissues.
In-vivo imaging of Aβ plaques in live mice with outstanding BBB penetrability. Inspired by the superb red uorescence light-up Aβ plaque-speci c response in the brain slices of mice, we further evaluated the biocompatibility and BBB penetrability of AIE-CNPy-AD which is indispensable before in-vivo imaging. Cell viability experiments were rst carried out to ascertain the biocompatibility of AIE-CNPy-AD by CCK-8 assays. Human neuroblastoma cells (SH-SY5Y), mouse brain neuroblastoma cells (Neuro-2a), mouse breast cancer cells (4T1) and human breast cancer cells (MCF-7) were incubated with different concentrations of AIE-CNPy-AD for 24 h, respectively. The cell viabilities of all these four cell lines both kept at a level close to 100% even at a probe concentration of 32 µM (Fig. S14A-Fig. S14D). The results provided strong evidence to the fact that AIE-CNPy-AD has low cytotoxicity and favorable biocompatibility to various cells, demonstrating its high applicability to live animals. The oil-water partition coe cient (log P) is often used as an index indicating the possible ability of penetrating BBB 45 . The log P of AIE-CNPy-AD was determined to be 1.24 by shaking-ask method, which is far larger than that of ThT (0.16), 8 suggestive of the higher lipophilicity and BBB penetrating potential of AIE-CNPy-AD as compared with ThT.
The feasibility of AIE-CNPy-AD tracking Aβ plaques in vivo was con rmed using 2.5-month-old 5*FAD mice, 6-month-old APP/PS1 mice and age-matched wild-type mice as model mice. Live imaging of these mice was performed after the tail vein injection of AIE-CNPy-AD (Fig. 5). Almost all the uorescence signals were clearly witnessed in the center of brain compartments and able to be e ciently captured. Apparently, the uorescence signals in the brain regions of 2.5-month-old 5*FAD mice (Fig. 5A) were already much stronger than those in the wild-type mice at 5 min post injection (Fig. 5B). Particularly, with the decay of signal, difference in signal intensity recorded from 5*FAD and wild-type mice was enlarged as indicated by the semi-quantitative analysis of the images (Fig. 5E). Similarly, intense uorescence signal is readily visible from the brain area of 6-month-old APP/PS1 mice after the injection of AIE-CNPy-AD. Moreover, the contrast between the uorescence signals from APP/PS1 and 6-month-old wild-type mice is quite dramatic. Compared with 5*FAD mice, difference in the signal intensity between APP/PS1 mice and age-matched wild-type mice was 4.7 times larger than that between 5*FAD mice and the corresponding wild-type mice at 5 min after being injected with AIE-CNPy-AD (Fig. 5G, 5H, and 5K). It might be because the formation of Aβ deposits in APP/PS1 mice is slower than that in 5*FAD mice. In addition, uorescence signal intensity of APP/PS1 mice declined faster than that of 5*FAD mice after being injected with AIE-CNPy-AD for 30 min, as it can be seen from the semi-quantitative analysis of the images (Fig. 5K). It is possibly because that there are larger and more Aβ plaques in 2.5-month-old 5*FAD mice than in 6-month-old APP/PS1 mice, which slows down the clearance of AIE-CNPy-AD from the brain.
Notably, whether in 5*FAD mice or APP/PS1 mice at 1 h post probe injection, F(t)/F(Pre) value of transgenic mice was at least 1.75 times that of wild-type mice, indicative of the potential of AIE-CNPy-AD to realize long-term tracking of Aβ plaques in vivo.
The case as for ThS is very different from that of AIE-CNPy-AD. As shown in Fig. 5C, 5D, 5F, 5I, 5J, and 5L, hardly any valid signal was found from the 5*FAD mice, APP/PS1 mice and wild-type mice administrated with ThS, clearly revealing the powerlessness of ThS in in-vivo imaging of Aβ plaques. This most probably results from the combination of its poor BBB permeability, short excitation and emission wavelengths which cannot penetrate the skull of mice, and the ACQ effect 46 . Undoubtedly, these visualization results directly validated that AIE-CNPy-AD is capable of penetrating the BBB and imaging Aβ brils/plaques in vivo with high contrast and delity.
In-vivo tracking of Aβ plaques in APP/PS1 mice at a very early stage. It has con rmed that AIE-CNPy-AD exhibits signi cant signal difference between transgenic mice and wild-type mice in a relatively long timeperiod post probe administration. It can be envisaged that AIE-CNPy-AD might be promising for early diagnosis of AD in transgenic mice by virtue of the Aβ plaques-sepeci c in-vivo imaging ability of AIE-CNPy-AD. Young APP/PS1 transgenic mice of 2-month-, 3-month-, 4-month-, and 6-month-old and agematched wild-type (WT) mice were thus employed to systematically assess the performance of AIE-CNPy-AD in early diagnosis of AD ( Fig. 6 and Fig. S15). It was observed that the difference in the uorescence signal intensity of APP/PS1 transgenic mice and wild-type mice already became very evident at the age of 4 month. In the meantime, the uorescence signal intensity is positively correlated with the age of transgenic AD mice (Fig. 6A). More speci cally, at 20 min post injection, the F(t)/F(Pre) value of the 4month-old APP/PS1 transgenic mice was 1.86 times that of the age-matched WT mice, and the F(t)/F(Pre) value of 6-month-old APP/PS1 transgenic mice was 2.06 times that of the age-matched WT mice (Fig. 6B). Meanwhile, there was no apparent signal difference in WT mice of different months (Fig.  6C), and the F(t)/F(Pre) values of 2-month-old and 3-month-old APP/PS1 transgenic mice were almost the same as those of the age-matched WT mice (Fig. 6D). These imaging results especially those acquired with APP/PS1 transgenic mice at an age of 4-month-old clearly showed that AIE-CNPy-AD is competent for the precise diagnosis of AD at a super-early stage. It has been speculated that the reason for the failure of early drug intervention on AD is probably that intervention in the phase is not early enough, which in turn is greatly related to the e cient capture of biological manifestations rather than clinical manifestations in the early diagnosis 47,48 . The experimental results obtained with AIE-CNPy-AD is of great signi cance because APP/PS1 transgenic mice are found to exhibit memory de cits from 5 months old on which is early clinical presentations, con rming that AIE-CNPy-AD can diagnose AD of APP/PS1 transgenic mice during the period of early biological manifestations, prior to the appearance of clinical presentations 43,47,49 . Furthermore, as shown in Fig. S16, ThS could neither give discriminatory signals between the APP/PS1 transgenic mice and wild-type mice nor provide distinct signals among mice of different ages, manifesting that our probe outperforms the clinically used gold-standard probe and could be used as an upgraded alternative to the commercially available ones.
It is worth mentioning that AIE-CNPy-AD does not only hold satisfactory cytocompatibility with negligible toxicity to diverse cells including neuronal cell lines but also has outstanding in-vivo biocompatibility as suggested by Fig. S17. The H&E staining results of the heart, liver, spleen, lung, kidney, and brain slices resected from 5*FAD (2-month-old), APP/PS1 (6-month-old) transgenic mice, and wild-type mice (6month-old) at 24 h post the administration with AIE-CNPy-AD via tail vein injection are almost identical to those injected with PBS buffer under parallel conditions. It manifests that AIE-CNPy-AD does not cause obvious necrosis to tissues, suggesting the superb in-vivo biocompatibility of the present probe. The Tunel results of the brain slices further veri es this as there is no evident apoptosis being observed in all the samples. Therefore, in view of the very low toxicity to cells and tissues, the AIE-CNPy-AD is proven to be highly suitable for in-vivo detection and has great potential in clinical use.
Ex-vivo observation of frozen brain slices of mice pre-administrated with AIE-CNPy-AD further validates its high delity of Aβ plaques-speci c imaging. To further examine whether the AIE-CNPy-AD binds exclusively to Aβ plaques in the brain of live mice, the APP/PS1 transgenic mice of different ages and age-matched wild-type mice were injected with AIE-CNPy-AD and sacri ced at 15 min post-administration. The frozen slices of these mice brain were then stained with rabbit anti-mouse primary antibody (ab201060), Alexa Fluor® 488-labeled goat anti-rabbit secondary antibody (ab150077), and Hoechst 33342 in sequence. As shown in Fig. 7A, 7F, and Fig. S17, the Aβ plaques were imaged with high resolution and high contrast. Moreover, it can be clearly observed that the Aβ plaques in brain slices of 6month-old APP/PS1 transgenic mice were larger and richer in number than those of 4-month-old APP/PS1 transgenic mic, while no signi cant Aβ plaques were found in the brain slices of 2-month-old, 3month-old APP/PS1 transgenic mice (Fig. S18), and all the age-matched WT mice (Fig. S19). Manifestly, AIE-CNPy-AD indeed labelled Aβ plaques in vivo with ultrahigh delity and speci city, as suggested by the very good overlap between the green (antibody) and red channels (AIE-CNPy-AD; Fig. 7B-7E and 7G-7J).
Elucidation of the working mechanism with molecular docking simulations. We believe that the rod-like architecture, extended D-π-A electronic structure, exible 3D conformation, the amphiphilic and zwitterionic molecular structure collectively contribute to the outstanding performance of AIE-CNPy-AD in Aβ brils/plaques-speci c detection and imaging. As above mentioned, the rod-like structure is supposed to favors the recognition of β sheets. Moreover, the hydrophobic π-conjugated backbone and the pyridyl and sulfonate groups are envisaged to bene t the binding of AIE-CNPy-AD to Aβ species possibly via hydrophobic interaction, π-π interaction, and electrostatic interaction. When in solution or molecularly dispersed, the AIE-CNPy-AD molecules undergo active intramolecular motions, leading to an OFF state and a low background. When coexists with Aβ monomer, intramolecular motions of AIE-CNPy-AD are merely weakly restricted, rendering weak red emission released (near-OFF state), due to the weak binding a nity of AIE-CNPy-AD and Aβ monomer. On the contrary, strong interactions between AIE-CNPy-AD and brils impose severe restriction on the intramolecular motions, which activates the AIE process 50,51 , and switches AIE-CNPy-AD from OFF to ON state to emit strong red/NIR uorescence. Thus, AIE-CNPy-AD can distinguish Aβ brils from Aβ monomers precisely (Fig. 8A).
Molecular docking simulations were carried out to unveil the interactions between AIE-CNPy-AD and Aβ brils or monomer and to elucidate the working mechanism of speci c detection of Aβ brils. As revealed by the three top docking conformations with the lowest binding free energies shown in Fig. 8B, 8C and Fig. S20, the most preferable binding direction of AIE-CNPy-AD to Aβ brils is consistent with the orientation of the β-sheets of Aβ brils, which proves that rod-shaped geometric structure of AIE-CNPy-AD is very helpful to the binding with Aβ brils 53 . Simultaneously, the hydrophobic PHE-19 residue on the βsheets of Aβ brils has strong C-H···π interaction (2.6 Å) with the phenyl ring of AIE-CNPy-AD, and the ALA-21 residue exhibits very strong interaction with the cyano group of AIE-CNPy-AD (2.2 Å). To our astonishment, the GLY-25 residue and two LYS-28 residues have strong hydrogen bonding interactions with the sulfonate group of AIE-CNPy-AD (1.7−2.1 Å). It is obvious that sulfonate not only improves water solubility of AIE-CNPy-AD, but also further enhances the binding ability of AIE-CNPy-AD to Aβ brils via Hbonding and electrostatic interactions. Collectively, the strong intermolecular interactions between AIE-CNPy-AD and Aβ brils greatly hinder the intramolecular motions of AIE-CNPy-AD to give out the "lightedup" uorescent response 54 .
In contrast, only the LYS-28 residue of Aβ monomer shows hydrogen bonding interaction with the sulfonate unit of AIE-CNPy-AD and no other interactions exist, which results in weaker restriction on the intramolecular motions and weak uorescence (Fig. 8D and Fig. 8E). The inhibition constant (K i ) 55 of AIE-CNPy-AD and Aβ brils (935 nM) is far smaller than that of AIE-CNPy-AD and Aβ monomer (45.0 μM).
Simultaneously, the lowest docking energy of AIE-CNPy-AD and Aβ brils was calculated to be -8. 23 kcal/mol, which is substantially lower than that of AIE-CNPy-AD and Aβ monomer (-5.93 kcal/mol). As exhibited in Table S1, all the ten best binding poses with the lowest energies of AIE-CNPy-AD and Aβ brils consistently display lower docking energies than those of AIE-CNPy-AD and Aβ brils. These data su ciently suggested that binding a nity of AIE-CNPy-AD to Aβ brils is much stronger than that of AIE-CNPy-AD to Aβ monomer. Moreover, the binding energy of AIE-CNPy-AD and Aβ brils is lower than that of ThT and Aβ brils (-7.18 kcal/mol) 56 , indicating that the AIE-CNPy-AD possesses higher a nity to Aβ brils as compared to ThT. The molecular docking simulation results are powerful proofs to our experiment results and further verify the rationality and feasibility of our design strategy.

Discussion
Precise in-vivo tracking of Aβ brils/plaques is of great signi cance in both fundamental research and technological development. In this work, we proposed an 'all-in-one' molecular design strategy and implemented it in the development of a novel AIE-active NIR-emissive probe (AIE-CNPy-AD) for speci cally discriminating Aβ brils and imaging Aβ plaques in vivo at an early stage. Rod-shaped amphipathic AIE-CNPy-AD possesses excellent BBB penetrability and high binding a nity to Aβ brils. The D-π-A electronic structure with multiple rotors and exible 3D conformations endues AIE-CNPy-AD with red/NIR emission, AIE feature, and resultant excellent in-vivo tracing capability of Aβ brils with superior tissue penetration ability, high signal-to-noise ratio and high delity. More importantly, our probe also enjoys the following advantages: (1) it is simple in structure, facile in preparation with cheap raw materials, and low in cost; (2) the in-vivo imaging capability is universal and not restricted by the strain of mouse; (3) the precise in-vivo tracking of Aβ brils could be realized at a very early stage before the occurrence of clinical manifestations; (4) the probe maps the Aβ plaques at the earliest stage as compared to all the reported uorescent probes; (5) the cytocompatibility and in-vivo compatibility is satisfactory for in-vivo investigation and clinical application (Table S2). Moreover, the rationality and feasibility were fully veri ed and the working mechanism of AIE-CNPy-AD was elucidated by the molecular docking simulation results. The present work will not only shed light on the rational design of high-performance uorescent probes for accurately imaging of Aβ brils in vivo, but also provide a promising diagnostic tool to diagnose AD at an early stage and nd timing of early drug intervention.

Methods
Materials and instruments. Conventional reagents and raw materials, purchased from formal channels, are analytically pure and used without further puri cation. Aβ 1-42 peptides were purchased from GL Biochem (Shanghai) Co., Ltd. Hen egg white lysozyme (HEWL) was purchased from Aladdin. Cell culture medium RPMI-1640, penicillin-streptomycin, and 0.25% trypsin were obtained from M&C Gene Technology (Beijing, China). Fetal bovine serum was purchased from Gemini Bio-Products (Calabasas, California, USA). All other solvents and reagents for biological experiments were of analytical grade.
The molecular structures were characterized using 1 H NMR, 13 C NMR and high-resolution mass spectroscopy. 1 H NMR and 13 C NMR spectra were recorded on a Brüker AV-400 spectrometer by using deuterated solvents with 0.03% TMS as internal standard. The high-resolution mass spectra were taken on an HP 5958 mass spectrometer with the electronic spray ionization mode. The UV-Vis absorption spectra were recorded in 1 cm slit quartz cells on an Agilent Cary 60 UV-Vis spectrometer. The photoluminescence spectra were recorded in 1 cm or 4 mm-slit quartz cells on an Agilent Cary Eclipse Fluorescence spectrophotometer. Photographs were taken with a Canon EOS 6D digital camera. Buffer solutions were prepared with Mettler Toledo FE28-Bio. Aβ brils samples were prepared by sonication in a SB-800DT water-bath sonicator and vortexed by digital Vortex-Genie® 2 mixer. TEM images were obtained on a JEM-1400 biological transmission electron microscope. Fluorescent images of brain slices were obtained by a confocal laser scanning microscope (CLSM, Nikon, Japan). The live mice were imaged with a LivingImage system and the relative uorescence intensity was analyzed by LivingImage 4.3.1 software (Caliper).
Synthesis of AIE-CNPy-AD. (Z)-3-(4-(Dimethylamino)phenyl)-2-(4-(pyridin-4-yl)phenyl)acrylonitrile (PyDPACN-N) is synthesized according to the literature. 57 1 (294 mg, 1.5 mmol), 2 (268 mg, 1.8 mmol), sodium hydroxide (72 mg, 1.8 mmol) and EtOH (20 mL) were added into a 100 mL two-necked roundbottomed ask. The ask was ushed with dry nitrogen for three times. Then the mixture was stirred overnight at room temperature. Yellow solids were ltered out and washed thoroughly with EtOH to yield 3 ( Fig. S1). and THF (10 mL) were added into a 100 mL two-necked round-bottomed ask. The ask was ushed with dry nitrogen for three times. Then the mixture was heated to 85 o C for 12 h. After the termination of the reaction, the resultant mixture was extracted with DCM for three times, and the collected organic layers was concentrated by a rotatory evaporator. Then the crude product was puri ed by column chromatography to obtain PyDPACN-N (Fig. S1).
PyDPACN-N (164 mg, 0.5 mmol), sodium 3-bromopropanesulfonate (5, 225 mg, 2.0 mmol), and CH 3 CN (20 mL) were added into a 100 mL two-necked round-bottomed ask. The ask was ushed with dry nitrogen three times. Then the mixture was re uxed for 96 h. Solids were ltered out from cooled mixture, and washed thoroughly with THF and water to afford the target product. was added into water to dilute to 10 μM solution. Samples for TEM measurement were prepared by Page 13/25 depositing 10 µL of the appropriate solution onto a carbon-coated copper grid, keeping at room temperature for 20 min to ensure the adsorption of Aβ 1-42 brils on the copper grid, and wicking the excess solution away with a small piece of lter paper. The samples were then stained with phosphotungstic acid. At last, the sample grids were allowed to be dried at room temperature and imaged on a JEM-1400 biological transmission electron microscope.
Cell lines for cytotoxicity assay. The human neuroblastoma cells (SH-SY5Y), mouse brain neuroblastoma cells (Neuro-2a), mouse breast cancer cells (4T1) and human breast cancer cells (MCF-7) were obtained from the Institute of Basic Medical Science (Beijing, China). SH-SY5Y cells were cultured in DMEM/F-12 medium. Neuro-2a cells were cultured in DMEM with high glucose medium. 4T1 and MCF-7 cells were cultured in RPMI-1640 medium. All of the above media were supplemented with 10% fetal bovine serum and 1 % (100 U/mL) penicillin-streptomycin at 37 o C in 5% CO 2 atmosphere.
In-vitro cytotoxicity assay. The cytotoxicity of AIE-CNPy-AD was investigated by Cell Counting Kit-8 (CCK-8) assay (Dojindo, Tokyo, Japan) according to the instruction manual. Brie y, human neuroblastoma cells (SH-SY5Y cells) were seeded into a 96-well plate at a density of 5000 cells/well for 24 h. Serial concentrations of AIE-CNPy-AD (0-32 μM, 200 μL/well) were added to treat cells. After being incubated for 24, culture medium with AIE-CNPy-AD was withdrawn, and then a fresh medium containing 10 μL of CCK-8 reagent was added to each well. Cells were incubated for another 3 h at 37 o C in 5% CO 2 atmosphere.
The absorbance at 450 nm was measured by a micro-plate reader (SpectraMAX 190, Molecular Devices). In-vitro fluorescence staining of mice brain slices using AIE-CNPy-AD. In-vitro uorescence staining of mice brain slices was performed to verify the binding ability of AIE-CNPy-AD to Aβ 1-42 aggregates. 5*FAD transgenic mice, APP/PS1 transgenic mice and age-matched wild-type mice were sacri ced with cold saline perfusion for 15 min without administration of AIE-CNPy-AD. Brain tissues were dissected, xed in 4% paraformaldehyde overnight, and dehydrated in 30% sucrose. Then the tissues were embedded with para n cut into 8 μm serial sections, and stored at -80 o C before use. Pre-prepared para n-embedded 4 μm brain tissue sections from 5*FAD transgenic mice (C57BL/6J, 2.5-month-old), APP/PS1 transgenic mice (C57BL/6J, 2, 3, 4, and 6-month-old) and age-matched wild-type mice (C57BL/6J, 2, 2.5, 3, 4, and 6month-old) were used for in-vitro uorescence staining. The tissue slices were baked in an incubator at 60 o C for 60 min to prevent stripping. After being cooled to room temperature, the slices were depara nized by immersion in xylene, followed by washing with different gradients of ethanol and water. Then the sections were heated to 95 o C in ethylene diamine tetraacetic acid (EDTA) antigen retrieval buffer for 20 min. After cooling, the sections were incubated in the aqueous solution of AIE-CNPy-AD (100 μM) at 37 o C for 20 min, and then washed with PBS buffer. The sections were blocked with 3% BSA solution at 37 o C for 60 min, permeabilized with 0.1% (v/v) Triton X-100 for 3 min at room temperature, then incubated with rabbit anti-mouse primary antibody (ab201060, diluted by 1:200 with PBS buffer) at 4 o C overnight.
Sections that were incubated only with 3% BSA were served as controls. After being washed 3 times with PBS buffer, all sections were stained with Alexa Fluor® 488-labeled goat anti-rabbit secondary antibody (ab150077, 1:1000) at 37 o C for 60 min. After being washed 3 times with PBS buffer, anti-uorescence quenching mounting solution containing Hoechst 33342 was added. Finally, samples were observed under a confocal laser scanning microscope (CLSM, Heidelberg, Germany), using the following conditions: Hoechst 33342: λ ex = 405 nm, λ em = 461 nm; Alexa Fluor® 488: λ ex = 495 nm, λ em = 519 nm; AIE-CNPy-AD: λ ex = 500 nm, λ em = 620 nm) Real-time in-vivo imaging in transgenic mice. The live imaging experiments were conducted to observe the distribution of AIE-CNPy-AD in brains. Before in-vivo imaging, the heads of transgenic mice of different ages and age-matched wild-type mice were shaved and cleaned. The background uorescence was captured by LivingImage system before the administration of AIE-CNPy-AD. Then AIE-CNPy-AD was injected to transgenic mice and age-matched wild-type mice (3 mice per group) via tail vein (2.0 mg kg -1 , 5.5% DMSO, 94.5% PBS buffer, 400 μL). In addition, the control groups of mice were injected with commercially available ThS. Then mice were anaesthetized by iso urane and photographed by LivingImage system at pre-set time points (AIE-CNPy-AD, λ ex = 500 nm, λ em = 620 nm; ThS, λ ex = 430 nm, λ em = 500 nm). The relative uorescence intensity was analyzed by LivingImage 4.3.1 software (Caliper).
Evaluation of the in-vivo biocompatibility of AIE-CNPy-AD. 5*FAD (2-month-old), APP/PS1 (6-month-old) transgenic mice, and wild-type mice (6-month-old) were chosen as models and were injected with AIE-CNPy-AD or PBS buffer via tail vein injection, then sacri ced with cold saline perfusion at 24 h after administration. Heart, liver, spleen, lung, kidney, and brain tissues were dissected from the above mice, xed in 4% paraformaldehyde overnight, and dehydrated in 30% sucrose. Then the tissues were embedded with para n cut into 8 μm serial sections, and stored at -80 o C before use. All the tissue slices were observed using hematoxylin-eosin staining. The brain slices were also observed utilizing TdTmediated DUTP Nick-End Labeling (Tunel) technique to detect the apoptosis.
Ex-vivo observation for uorescence signals of the brain slices of mice intravenously injected with AIE-CNPy-AD. APP/PS1 transgenic mice of different ages and age-matched wild-type mice were injected with AIE-CNPy-AD, then sacri ced with cold saline perfusion at 15 min after administration. Brain tissues were dissected, xed in 4% paraformaldehyde overnight, and dehydrated in 30% sucrose. Then the tissues were embedded with OCT (optimum cutting temperature embedding medium), cut into 8 μm serial sections, and stored at -80 o C before use. The sections were blocked with 3% BSA solution at 37°C for 60 min, permeabilized with 0.1% (v/v) Triton X-100 for 3 min at room temperature, then incubated with rabbit anti-mouse primary antibody overnight at 4°C. After being washed 3 times with PBS buffer, all sections were stained with Alexa Fluor® 488-labeled goat anti-rabbit secondary antibody at 37 °C for 60 min, followed by the addition of anti-uorescence quenching mounting solution containing Hoechst 33342. Finally, samples were observed with a confocal laser scanning microscope (CLSM, Heidelberg, Germany).
Docking simulations and analysis. The structure of AIE-CNPy-AD was optimized with a B3LYP/6-31+G* basis set using the Gaussian 09 package 60