Rationally designed divalent caffeic amides inhibit amyloid-β fibrillization, induce fibril dissociation, and ameliorate cytotoxicity

Article abstract of DOI:10.1016/j.ejmech.2018.08.084

One of the pathologic hallmarks in Alzheimer's disease (AD) is extracellular senile plaques composed of amyloid-β (Aβ) fibrils. Blocking Aβ self-assembly or disassembling Aβ aggregates by small molecules would be potential therapeutic strategies to treat AD. In this study, we synthesized a series of rationally designed divalent compounds and examined their effects on Aβ fibrillization. A divalent amide (2) derived from two molecules of caffeic acid with a propylenediamine linker of ～5.0 ? in length, which is close to the distance of adjacent β sheets in Aβ fibrils, showed good potency to inhibit Aβ(1–42) fibrillization. Furthermore, compound 2 effectively dissociated the Aβ(1–42) preformed fibrils. The cytotoxicity induced by Aβ(1–42) aggregates in human neuroblastoma was reduced in the presence of 2, and feeding 2 to Aβ transgenic C. elegans rescued the paralysis phenotype. In addition, the binding and stoichiometry of 2 to Aβ(1–40) were demonstrated by using electrospray ionization?traveling wave ion mobility?mass spectrometry, while molecular dynamic simulation was conducted to gain structural insights into the Aβ(1–40)?2 complex.

Full text of DOI:10.1016/j.ejmech.2018.08.084

Accepted Manuscript
Rationally designed divalent caffeic amides inhibit amyloid-β fibrillization, induce fibril
dissociation, and ameliorate cytotoxicity
Ling-Hsien Tu, Ning-Hsuan Tseng, Ya-Ru Tsai, Tien-Wei Lin, Yi-Wei Lo, Jien-Lin
Charng, Hua-Ting Hsu, Yu-Sheng Chen, Rong-Jie Chen, Ying-Ta Wu, Yi-Tsu Chan,
Chang-Shi Chen, Jim-Min Fang, Yun-Ru Chen
PII:
S0223-5234(18)30755-4
10.1016/j.ejmech.2018.08.084
EJMECH 10695
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 31 December 2017
Revised Date: 28 August 2018
Accepted Date: 28 August 2018
Please cite this article as: L.-H. Tu, N.-H. Tseng, Y.-R. Tsai, T.-W. Lin, Y.-W. Lo, J.-L. Charng, H.-T.
Hsu, Y.-S. Chen, R.-J. Chen, Y.-T. Wu, Y.-T. Chan, C.-S. Chen, J.-M. Fang, Y.-R. Chen, Rationally
designed divalent caffeic amides inhibit amyloid-β fibrillization, induce fibril dissociation, and ameliorate
cytotoxicity, European Journal of Medicinal Chemistry (2018), doi: 10.1016/j.ejmech.2018.08.084.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Rationally Designed Divalent Caffeic Amides Inhibit Amyloid-β Fibrillization, Induce
Fibril Dissociation, and Ameliorate Cytotoxicity
1
, †,‡
1,†
2,†
1
1
Ling-Hsien Tu
, Ning-Hsuan Tseng , Ya-Ru Tsai , Tien-Wei Lin , Yi-Wei Lo , Jien-Lin
1
1
2
1
1
2
Charng , Hua-Ting Hsu , Yu-Sheng Chen , Rong-Jie Chen , Ying-Ta Wu , Yi-Tsu Chan ,
3
1,2,
1,
Chang-Shi Chen , Jim-Min Fang * and Yun-Ru Chen *
1
Genomics Research Center, Academia Sinica, Taipei 115, Taiwan.
2
Department of Chemistry, National Taiwan University, Taipei 106, Taiwan.
3
Department of Biochemistry and Molecular Biology, National Cheng Kung University,
Tainan, Taiwan.
†
These authors contributed equally to this work.
‡
Current address of L.-H.T: Department of Chemistry, National Taiwan Normal University,
Taipei 106, Taiwan.
*
To whom correspondence should be addressed: Jim-Min Fang, E-Mail: jmfang@ntu.edu.tw;
Yun-Ru Chen, E-Mail: yrchen@gate.sinica.edu.tw
1

ACCEPTED MANUSCRIPT
Abstract
One of the pathologic hallmarks in Alzheimer’s disease (AD) is extracellular senile
plaques composed of amyloid-β (Aβ) fibrils. Blocking Aβ self-assembly or disassembling Aβ
aggregates by small molecules would be potential therapeutic strategies to treat AD. In this
study, we synthesized a series of rationally designed divalent compounds and examined their
effects on Aβ fibrillization. A divalent amide (2) derived from two molecules of caffeic acid
with a propylenediamine linker of ~5.0 Å in length, which is close to the distance of adjacent
β sheets in Aβ fibrils, showed good potency to inhibit Aβ(1-42) fibrillization. Furthermore,
compound 2 effectively dissociated the Aβ(1-42) preformed fibrils. The cytotoxicity induced
by Aβ(1-42) aggregates in human neuroblastoma was reduced in the presence of 2, and
feeding 2 to Aβ transgenic C. elegans rescued the paralysis phenotype. In addition, the
binding and stoichiometry of 2 to Aβ(1-40) were demonstrated by using electrospray
ionization−traveling wave ion mobility−mass spectrometry, while molecular dynamic
simulation was conducted to gain structural insights into the Aβ(1-40)−2 complex.
2

ACCEPTED MANUSCRIPT
Keyword
Amyloid-β; Alzheimer’s disease; caffeic acid; divalent caffeic amide; inhibitor; cytotoxicity.
Research highlights
ó
ó
ó
ó
ó
Compounds bearing two units of caffeic acid with different linkers are synthesized.
A divalent caffeic amide (2) shows potent inhibition against Aβ fibrillization.
ESI-TWIM-MS reveals the binding of compound 2 with Aβ monomer and oligomer.
Compound 2 disassembles Aβ fibrils and ameliorates Aβ induced cytotoxicity.
Compound 2 extends lifespan of Aβ transgenic C. elegans, potential for AD therapy.
Abbreviation
Aβ, amyloid-β; AD, Alzheimer’s disease; CA, caffeic acid; CCS, collision cross-section; DMF,
dimethylformamide;
EDCI,
1-ethyl-3-(3-dimethylaminopropyl)
carbodiimide;
ESI-TWIM-MS, electrospray ionization-traveling wave ion mobility-mass spectrometry;
HFIP, hexafluoroisopropanol; HOBt, 1-hydroxybenzotriazole; HSQC, heteronuclear single
quantum coherence; SEM, standard error of mean; NGM, Nematode Growth Media; PyBop,
3

ACCEPTED MANUSCRIPT
benzotriazol-1-yl-oxytripyrrolidinophosphonium; TEM, transmission electron microscopy;
ThT, thioflavin T.
4

ACCEPTED MANUSCRIPT
1
. Introduction
Aβ peptide composed of 40 or 42 residues is generated from proteolysis of amyloid
precursor protein by β- and γ-secretases. Aβ aggregation is highly implicated in Alzheimer’s
disease (AD) pathogenesis [1]. Aβ is an intrinsically disordered and highly aggregation-prone
peptide. Aβ fibrillization starts with a lag phase in which the protein monomers gradually
associate into soluble oligomers, and followed by a rapid elongation phase of fibril formation
to reach a steady phase where mature amyloid fibrils are formed. The structure of Aβ fibrils
has been revealed at different levels. Aβ(1-42) fibrils consist of three parallel β-sheets (i.e. aa
1
2-18, 24-33, and 36-40) [2, 3], which are different from the Aβ(1-40) fibril comprising two
parallel β-sheets (i.e. aa 10-22 and 30-40) [4]. The distance of ~4.7 Å of intermolecular
in-register β-sheets derived from the measurement of two adjacent carbonyl groups by
solid-state nuclear magnetic resonance (NMR) is consistent with a typical cross-β diffraction
pattern revealed by X-ray fiber diffraction [5]. The amyloid positron-emission tomography
(PET) tracers are currently used for probing the fibril amount in brain of the patients to
monitor the progression of Alzheimer’s disease [6]. The process of Aβ aggregation has been
suggested to induce synaptic dysfunction and cause neurotoxicity [7]. Even though Aβ(1-40)
5

ACCEPTED MANUSCRIPT
is more abundant than Aβ(1-42) in AD patient’s brain, Aβ(1-42) aggregates more aggressively
and exhibits markedly higher toxicity [8, 9]. Since Aβ aggregation is apparently a key target
for AD treatment, many researchers aim to develop therapeutic approaches that prevent
production of Aβ peptides or inhibit their abnormal aggregation [10−12].
Screening or designing small drug-like molecules targeting Aβ has been considered as
one of the promising strategies for therapeutics of AD [13]. The concepts include that small
molecules disrupt protein−protein interactions among Aβ monomers, stop conformational
change of Aβ from its intermediates to β-sheet fibrillar state, and indirectly inhibit Aβ
aggregation by sequestering excess metals [14]. Previous studies have identified small
molecules such as Congo red [15], chrysamine G [16], and curcumin [17−22] as potent
inhibitors of Aβ. These inhibitors share a similar chemical motif containing two substituted
aromatic groups to bind with Aβ. It suggests that an ideal Aβ inhibitor should contain terminal
groups interacting with the residues in or adjacent to the segments that are considered crucial
for peptide aggregation. Thus, the linker should have an appropriate length to join the two
terminal aromatic groups spanning the segments. In another aspect, caffeic acid (CA), a
chemical constituent widely distributed in edible plants, is also found to protect neuronal cells
6

ACCEPTED MANUSCRIPT
from Aβ-induced cytotoxicity [23], presumably because the catechol moiety can provide
hydrogen bonding and hydrophobic interaction with Aβ peptide. Inspired by these findings,
we thus designed and synthesized a series of divalent compounds based on the scaffold of CA
to examine their effects on Aβ fibrillization.
2
. Results
2
.1 Design and synthesis of divalent compounds.
To take advantage of multivalent effect on enhancement of binding affinity [24], the
divalent compounds 1−11 (Fig. 1) were designed as possible inhibitors against formation of
Aβ fibrils. Five dimeric CA amides 2−6 are equipped with different lengths of straight-chain
linkers. The linkers include propylenediamine (a short linker of ~5.0 Å length) and a series of
ethylene glycol diamines (up to ~22.5 Å length). The linkers are chosen for easy linkage to
two CA motifs, thus the prepared dimeric compounds would have good solubility in PBS
buffer. In comparison, the dimeric CA molecules 7−10 have more rigid linkers by
incorporation of phenyl and triazole rings. CA is a catechol-containing molecule, whereas
curcumin molecule contains monomethylated catechol rings. Compound 1 (R = H), the
7

ACCEPTED MANUSCRIPT
demethylated derivative of curcumin (12, R = Me), is also designed because it shares the
structural feature similar to that of divalent caffeic compound. Rosmarinic acid (RA) is a
natural compound structurally formed by addition of CA to the double bond of another CA
molecule. It would be interesting to synthesize a dimeric RA compound 11 and examine its
binding potency with Aβ.
Fig. 1. Structures of divalent compounds derived from the scaffold of caffeic acid. The length
of diamine linker and the distance between two catechol groups in compound 2 were
estimated by ChemDraw 15.0 software (PerkinElmer).
Scheme 1 shows the synthetic pathways to the divalent compounds 1−11. Demethylation
of curcumin was readily accomplished by treatment with Lewis acid BBr to give compound
3
8

ACCEPTED MANUSCRIPT
1
[25].
a
Scheme 1. Synthesis of diva lent compounds 1−11.
a
o
Reagents and reaction conditions: (a) BBr , CH Cl , −78 C (15 min), rt (12 h); 63%. (b)
3
2
2
EDCI, HOBt, Et N, DMF, rt, 18 h; 17, 93%; 18, 71%; 19, 51%. (c) Ac O, py, rt, 12 h. (d)
3
2
K CO , MeOH, rt, 0.5 h; 2, 27%; 3, 58%; 4, 58%; 5, 36%; 6, 34%. (e) CuSO , sodium
2
3
4
o
ascorbate, THF/H O (1:1), 40 C, 12 h; 7, 67%; 8, 26%; 9, 59%; 10, 20%; 11, 52%. (f) PyBop,
2
Et N, DMF, rt, 18 h; 62%.
3
The coupling reaction of diamine linker with two molecules of CA was carried out by the
promotion
of
1-ethyl-3-(3-dimethylaminopropyl)
carbodiimide
(EDCI)
and
1
-hydroxybenzotriazole (HOBt) in N,N-dimethyl formamide (DMF) solution. The crude
coupling products were subjected to acetylation, followed by deprotection of the acetyl
9

ACCEPTED MANUSCRIPT
groups, to facilitate isolation of the dimeric CA compounds 2−6. Alternatively, the amines
1
4−16 containing a terminal alkynyl group were reacted with CA to afford amides 17−19,
which underwent a Cu(I)-catalyzed (3+2) dipolar cycloaddition [26] with bis-azides 20 or 21
to furnish the triazole formation, giving compounds 7−10. The coupling reaction between
rosmarinic acid
and amine 14
was
realized
by the
promotion of
benzotriazol-1-yl-oxytripyrrolidinophosphonium (PyBop) to afford compound 23, which was
subsequently treated with linker 21 to give the dimeric RA compound 11 via click reaction
[
26].
2
.2 Inhibition of Aβ fibrillization.
Thioflavin T (ThT) assay was employed to monitor the inhibitory effect of the synthetic
dimeric compounds on Aβ aggregation. ThT is a classic amyloid dye that binds to crossed
β-sheets of amyloid fibrils and emits fluorescence upon fibril formation [27]. The percentage
of fibril formation corresponding to the presence of individual test compound was calculated
from the relative intensity of fluorescence at 485 nm, compared with 100% fibril formation in
phosphate buffer (pH 7.4). Our initial screening of compounds 1−11 indicated that the dimeric
compound 2 was an effective inhibitor at a concentration of 10 µM, rendering about 50%
1
0

ACCEPTED MANUSCRIPT
reduction of Aβ(1-40) fibrillization (Table 1).
a
Table 1. Effect of test compounds against fibril formation of Aβ(1-40).
a
a
compound fibril formation (%) compound fibril formation (%)
b
1
2
3
4
5
6
7
8
9
11.5 ± 0.4
52.9 ± 2.5
64.7 ± 5.6
79.4 ± 2.5
61.8 ± 2.0
67.6 ± 4.1
76.5 ± 7.8
88.2 ± 1.5
111.8 ± 2.5
10
11
13
17
18
19
22
23
61.8 ± 2.5
65.1 ± 10.6
91.2 ± 9.1
79.4 ± 6.1
135.3 ± 7.6
67.6 ± 6.1
79.1 ± 7.6
107.0 ± 1.0
a
Aβ(1-40) of 25 µM in PBS (pH 7.4) was incubated with test compound (10 µM), and the
fibril formation was monitored by ThT assay. The percentage of fibril formation was
calculated from the relative intensity of fluorescence emission at 485 nm by excitation at
442 nm wavelength. Control is PBS, corresponding to 100% fibril formation in the absence
of test compound. The experiments were performed in duplicate.
b
The percentage of fibril formation was likely underestimated because compound 1 (the
1
1

ACCEPTED MANUSCRIPT
demethylated derivative of curcumin) in PBS had UV absorption band at λmax = 437 nm (ε =
−1
−1
2
600 M cm ), which would interfere with ThT assay using excitation wavelength of 442
nm.
The inhibitory activities of divalent caffeic amides 3−10 varied depending on the types
and lengths of linkers. The one-arm compounds, such as caffeic acid (13) and its amide
derivatives 17−19, were less effective inhibitors. Rosmarinic acid (22) showed a relatively
low inhibitory activity [20], and its amide derivative 23 was inactive at 10 µM. Nonetheless,
the Aβ inhibitory activity was enhanced in dimeric RA compound (11). Demethylated
curcumin (1) in PBS solution likely exists as the enol tautomer to have an absorption band at
λmax = 437 nm, which would interfere with the ThT assay using 442 nm light for excitation.
Thus, the real inhibitory activity of 1 against Aβ fibril formation could not be determined by
this ThT assay.
Since Aβ(1-42) is a more aggressive and toxic Aβ isoform than Aβ(1-40), we
investigated the efficiency of curcumin (12), CA (13) and dimeric CA (diCA, 2) on preventing
Aβ(1-42) fibrillization. The solution of Aβ(1-42) at 50 µM with 5 µM ThT was co-incubated
o
individually with 100, 25, 12.5, 3.1, 0.8, 0.1 and 0.01 µM of test compound at 37 C. The ThT
intensity was monitored for a period over 80 h. We found that all three compounds could
1
2

ACCEPTED MANUSCRIPT
inhibit the fibrillization of Aβ(1-42) in a dose-dependent manner (Fig. 2a−2c). The
normalized final ThT intensity (at 80 h incubation) was plotted against compound
concentration and the data were fitted to obtain half maximal inhibitory concentration (IC )
5
0
for compounds 2 (IC = 4.9 ± 0.08 µM), 12 (IC = 3.4 ± 0.07 µM) and 13 (IC = 15.6 ±
5
0
50
50
0
.13 µM) (Fig. 2d).
Fig. 2. The inhibitory effect of curcumin (12), caffeic acid (13), and divalent caffeic amide
diCA, 2) on Aβ(1-42) fibrillization. Fibrillization of Aβ(1-42) in the presence of (a) 12, (b)
(
1
3, and (c) 2 were monitored by ThT assay. The fluorescence intensity was recorded (λ =
ex
4
42 nm, λ = 485 nm). Experiments in triplicate were performed in 20 mM phosphate buffer
em
o
(
pH 7.4) at 37 C containing 50 µM Aβ(1-42) in the absence or presence of 0.01, 0.1, 0.8, 3.1,
1
2.5, 25, and 100 µM compounds. Mean value and standard error of mean (SEM) were shown
for each time point. (d) The percentage of inhibition calculated from the final normalized ThT
signals was plotted against compound concentration. The data were fit to an equation Y =
1
00/ (1+10^(X-LogIC )) to calculate IC values. SEM was shown; however, some are too
50 50
little to be seen.
1
3

ACCEPTED MANUSCRIPT
Next, morphology of the end-point products after 80 h incubation was examined by
transmission electron microscopy (TEM). In the absence of test compound, the Aβ(1-42)
sample formed clustered and dense fibrils (Fig. 3a). In contrast, when Aβ(1-42) was
co-incubated with 25 µM of test compound, no fibril was found and only some amorphous
aggregates appeared (Fig. 3b−3d). Also, when Aβ(1-42) was co-incubated with 12.5 µM of
test compound, only a few fibrils and short filaments were observed (Fig. 3e−g).
Fig. 3. TEM images of 50 µM Aβ(1-42) after incubation in the absence and presence of test
compounds. Aβ(1-42) in the absence of test compound (a) and in the presence of 25 µM
curcumin (b), CA (c), and diCA (d). Aβ(1-42) in the presence of 12.5 µM curcumin (e), CA
(f), and diCA (g). The scale bar represents 100 nm.
To find the influence of linker length in inhibition of Aβ(1-42) fibrillization, we
1
4

ACCEPTED MANUSCRIPT
compared the binding affinity of dimeric CA amides 2−6. Aβ(1-42) was first prepared at 50
µM in the presence of ThT, and then co-incubated individually with test compound at 0.8, 3.1,
o
1
2.5, 25 and 100 µM at 37 C. The ThT intensity was normalized with that of Aβ(1-42) alone
(Fig. S1a in Supporting Information). Although all the CA dimers can affect Aβ(1-42)
fibrillization with increasing concentration, compound 2 with ~5.0 Å linker greatly reduced
the final ThT intensity, suggesting that compound 2 has good potential to inhibit Aβ(1-42)
fibril formation. When 50 µM Aβ(1-42) was incubated with 3.1 µM of 2, the final ThT
intensity of Aβ(1-42) with the compound was reduced to ~60% compared with that of Aβ
fibrils only. With increase of the inhibitor 2 to 25 and 100 µM concentrations, the final ThT
intensity significantly reduced to ~8% and ~1%, respectively. We also confirmed the ThT
results by taking TEM images. All the images were recorded for Aβ with and without 12.5
µM CA dimers. The result showed that all Aβ(1-42) peptides formed long and extensive
fibrils in the absence (supplementary Fig. S1b) and presence of most CA dimers
(supplementary Fig. S1d−g) except for compound 2 (supplementary Fig. S1c). Incubation of
Aβ(1-42) with compound 2 only led to formation of some short filaments.
1
5

ACCEPTED MANUSCRIPT
2
.3 Compound 2 disassembles Aβ fibrils.
In addition, we examined the ability of diCA (2) in disassembling preformed Aβ fibrils.
Aβ(1-42) fibrils were first prepared at 25 µM. Then, diCA at different concentrations, ranging
from 0 to 100 µM, was added into the preformed fibrils with 5 µM ThT, and the change of
ThT fluorescence was monitored with buffer background subtracted (Fig. 4a). We found
Aβ(1-42) fibrils alone have some decrease in ThT intensity which may be due to gradual
sedimentation of fibrils and/or clustering of Aβ(1-42) fibrils after a period of time.
Interestingly, we found that diCA is able to reduce mature fibrils in a dose-dependent manner
with IC50 of 4.2 ± 0.03 µM (Fig. 4b). Since fluorescence signal may be affected by
environment and binding mode, we further confirmed the Aβ assembly by loading the
samples in the absence and presence of different concentrations of 2 (25, 50 and 100 µM) at
time 0 and after 14 days of incubation to native Tris-tricine gel and subjected to western blot
(Fig. 4c) and TEM (Fig. 4d and 4e) examination. The western blot results showed that there is
a slight reduction of aggregates in the presence of different concentrations of 2 at time 0.
After 14 days of incubation with 2 there are much less or even no aggregates on the gel. We
also confirmed that little filaments were observed when Aβ(1-42) fibrils were incubated with
1
6

ACCEPTED MANUSCRIPT
1
00 µM of 2 after 14 days in TEM images (Fig. 4e) whereas Aβ(1-42) fibrils with 100 µM of
2
at time 0 still remained at least in some degree as dense fibrils (Fig. 4d). The result showed
that the CA dimer 2 reduced the preformed Aβ fibrils as evidence by reduced ThT intensity,
reduced aggregates on the top of western blots, and reduced fibrils in TEM images.
Fig. 4. Compound 2 (diCA) disassembles preformed Aβ(1-42) fibrils. (a) Preformed Aβ(1-42)
fibrils were incubated with 0 to 100 µM diCA solution containing 1% DMSO and 5 µM ThT.
The assay was performed in triplicate. (b) The final ThT fluorescence intensity (buffer
background subtracted) in the presence of diCA was normalized to Aβ(1-42) fibrils alone and
plotted with the log value of compound concentration. The data in three sets were fit to an
equation Y = 100/ (1+10^ (X-LogIC )) to calculate IC value. (c) Aβ(1-42) fibrils with 0, 25,
5
0
50
5
0 and 100 µM diCA were incubated for 0 and 14 days, and the aggregates were visualized by
gel electrophoresis with western blotting probed by anti-Aβ antibody 4G8 and 6E10. (d) TEM
image of Aβ(1-42) fibrils with 100 µM diCA without incubation. (e) TEM image of Aβ(1-42)
fibrils with 100 µM diCA after 14 days incubation. The scale bar represents 100 nm.
1
7

ACCEPTED MANUSCRIPT
2
.4 Compound 2 binds to Aβ monomer and oligomers.
To understand the molecular interaction of CA dimer 2 with Aβ, we utilized electrospray
ionization−traveling wave ion mobility−mass spectrometry (ESI-TWIM-MS) to detect the
possible complexes formed by Aβ with compound 2. ESI-TWIM-MS has been widely utilized
in differentiating isomeric structures with similar sizes [30, 31]. By applying voltage pulses in
the ion-mobility cell, the ions are propelled to drift against a stream of buffer gas [32, 33].
The time that ions need to drift through the ion-mobility cell is dependent on their m/z ratios
and molecular shape. At the same m/z ratio, the more compact ion reveals a shorter drift time
so that isomer differentiation can be realized. This technique has been frequently utilized in
characterizing the ion species in a complex mixture. Thus, one can use ESI-TWIM-MS to
examine the binding capacity of ligands to target protein. Recently, several binding modes
classified from a series of small molecules inhibitors against Aβ(1-40) and islet amyloid
polypeptide indicated that the potent inhibitors usually can bind protein in a stoichiometric
manner [34]. Due to fast aggregation of Aβ(1-42) that would complicate the ESI-TWIM-MS
experiment, Aβ(1-40) is commonly used to investigate its molecular interactions with ligands.
In our study, Aβ(1-40) alone at 35 µM or with 200 µM of compound 2 in a 1:5.7 molar ratio
1
8

ACCEPTED MANUSCRIPT
(Aβ/2) were prepared and injected for recording the ESI-TWIM-MS spectra. The experiments
revealed that Aβ(1-40) yielded dominantly the 3+ and 4+ monomeric ions as well as the 4+
and 5+ dimeric ions, appearing at m/z 1444, 1083, 2166, and 1733, respectively (Table 2).
The result is consistent with the previous literature [34]. When diCA (2) was present in the
Aβ(1-40) solution, one to four diCA molecules were found to bind with 3+ ion of Aβ(1-40)
monomer and one to three diCA molecules were detected to be with 4+ and 5+ ions of
Aβ(1-40) dimer (Fig. 5 and Table 2). The 3+ ions of Aβ−2 complexes at 1:1, 1:2 and 1:3
stoichiometry which occurred at m/z = 1576, 1709, and 1842, respectively, were indicated by
orange stars in Fig. 5a and 5b. In addition, Aβ dimer and their diCA complexes of 2:1 and 2:2
stoichiometry in both 4+ and 5+ states were also resolved (Table 2), indicating that the
binding mode of diCA to Aβ(1-40) monomer or dimer is specific. The collision cross-section
(CCS) value of each complex was calculated and listed in Table 2.
1
9

ACCEPTED MANUSCRIPT
Table 2. Experimental mass to charge ratio, drift time and collision cross-section (CCS) of
Aβ(1-40) and its complexation species with diCA (2).
a
2
species
M
charge state
m/z
drift time (ms)
9.81
CCS (Å )
4+
3+
3+
3+
3+
3+
5+
5+
5+
5+
4+
4+
4+
1083.2816
1444.0408
1576.7568
1709.8217
1842.5073
1975.2498
1733.0510
1812.6821
1892.5221
1971.7124
2165.8137
2265.1050
2365.1323
695.2
602.9
633.8
668.4
699.3
727.3
973.3
991.0
1008.4
1042.9
952.4
977.0
1001.2
M
12.24
13.23
14.33
15.44
16.43
11.69
M+L
M+2L
M+3L
M+4L
2
M
2
M+L
12.02
12.35
13.01
15.99
16.65
17.31
2
2
M+2L
M+3L
2
M
2
M+L
2
M+2L
2
0

ACCEPTED MANUSCRIPT
a
M represents Aβ(1-40); 2M represents dimer of Aβ(1-40); L represents diCA (2).
Fig. 5. ESI-TWIM-MS and MD simulation of the complexation species of Aβ(1-40) with
diCA (2). (a) ESI-TWIM-MS plot of 35 µM Aβ(1-40) with 200 µM diCA in 50 mM
ammonium acetate (pH 6.8). DiCA binds to the 3+ ion of Aβ monomer and to the 4+ and 5+
2
1

ACCEPTED MANUSCRIPT
ions of Aβ dimer. Aβ and diCA complexes are indicated by orange stars, and the number of
orange stars represents the number of diCA bound to Aβ. (b) The positive-mode ESI mass
spectrum of Aβ(1-40) with diCA. Monomeric Aβ and its diCA complexes are shown. (c) & (d)
represent the two groups of diCA and Aβ complex structures from MD simulation. The side
chain of first amino acid residue in Aβ(1-40) is shown. The calculated CCSs match our
experimental data.
In an attempt to elucidate the structural basis for binding of Aβ with diCA (2) by nuclear
1
15
magnetic resonance (NMR) spectroscopy [28, 29], we acquired the H- N heteronuclear
single quantum coherence (HSQC) spectrum (supplementary Fig. S2). Despite compound 2 is
a good inhibitor against Aβ fibrillization, we did not observe apparent change of chemical
1
5
shifts from cross peaks of N-labeled Aβ(1-42) at 25 µM when it was incubated with 2 at 500
µM. The reason for insignificant change in the HSQC spectrum is unclear; however, similar
phenomena have been shown in other inhibition studies [28, 29]. One possible explanation is
that the interaction between compound 2 and Aβ(1−42) is transient and may be at a relatively
low concentration that cannot effectively induce NMR chemical shifts. No further NMR
experiment was attempted in this study.
Alternatively, we employed computational methods to obtain the structural insights into
the complexation of Aβ monomer with ligand 2. Since Aβ monomer is natively disordered,
2
2

ACCEPTED MANUSCRIPT
the only available PDB file for Aβ(1-40) was determined by NMR (PDB code 2LFM) and
used for molecular docking experiment. According to our docking results, 9 dominant
conformations of Aβ−2 complex were revealed. The small molecule 2 seems to preferentially
interact with Aβ near the α-helical region and disrupt the partially folded helix structure. We
also conducted molecular dynamic (MD) simulations based on 11 docked structures of Aβ−2
complex to identify the possible conformers observed in our ESI-TWIM-MS experiments.
From MD trajectory, a hundred structures were sampled from each starting complex. Each
conformation was converted to the corresponding CCS by the trajectory method (TM) in
MOBCAL [35]. Their optimized potential force field energy and individual orientationally
averaged CCS values were also calculated (Fig. S3). Comparing the CCS values in
ESI-TWIM-MS experiment and MD simulation, most of the calculated CCS values for the
2
complex derived from MD simulation are larger than our experimental data (e.g., 634 Å for a
3
+ state of Aβ−2 complex in Table 2). However, we still found several similar conformers
whose CCS values matched well to our experimental data. Hence, two groups of
representative Aβ−2 complex are depicted in Figure 5c and 5d.
2
3

ACCEPTED MANUSCRIPT
2
.5 Compound 2 ameliorates Aβ(1-42) induced cytotoxicity.
To examine possible rescue effect of diCA (2), the end-point products of Aβ(1-42) in
ThT assay after incubation with and without test compound at different concentrations were
treated to human neuroblastoma SH-SY5Y cells and subjected to cytotoxicity assay. After 48
h incubation, cytotoxicity was assessed by lactate dehydrogenase (LDH) assay (Fig. 6a). The
LDH assay showed that Aβ(1-42) fibrils induced ~49% cytotoxicity and this deleterious effect
was gradually attenuated after co-incubating with curcumin (12), CA (13) or diCA (2) in a
dose dependent manner. Among all concentration tested, diCA is the most potent compound
in comparison to the others. At 0.01 µM, diCA was able to reduce cytotoxicity from 49% to
3
0% (rescue ~19%), whereas CA was only able to rescue ~5% and curcumin has no effect at
this concentration. At 25 µM, diCA was able to reduce cytotoxicity from 49% to 15% (rescue
~34%), whereas CA and curcumin were able to reduce cytotoxicity from 49% to ~27%
(rescue ~22%). Although all three compounds were able to rescue Aβ(1-42) induced toxicity,
diCA provided the most beneficial effect even at low concentration.
2
4

ACCEPTED MANUSCRIPT
Fig. 6. Rescue effect of test compounds against Aβ(1-42) induced toxicity. (a) Cytotoxicity of
Aβ(1-42) in the absence and presence of curcumin (12), CA (13), and diCA (2). The end
products of ThT assay were subjected to SH-SY5Y cells and the cytotoxicity was determined
by LDH assay. The experiments were performed in triplicate. The data were normalized to the
positive control, i.e. cells treated with Triton X-100. The statistical significance was indicated
by two-way ANOVA (***, p < 0.0001; **, p < 0.005; *, p < 0.05). (b) Cytotoxicity of
preformed Aβ(1-42) fibrils in the presence of diCA. The samples in the absence and presence
of diCA at different concentrations were treated to neuroblastoma SH-SY5Y cells.
Cytotoxicity was measured by LDH assay. Cells treated with Triton X-100 were used as the
positive control for 100% cytotoxicity. The experiments were performed in triplicate. The
statistical significance was indicated by one-way ANOVA (***, p < 0.0001).
To understand whether diCA could rescue the cytotoxicity induced by preformed Aβ
fibrils, we first prepared Aβ(1-42) fibrils in the absence and presence of diCA at different
concentrations (0, 25, 50 and 100 µM). The samples were then treated to SH-SY5Y cells and
subjected to LDH assay (Fig. 6b). The result showed that Aβ(1-42) fibrils alone caused
cytotoxicity about 17%, while fibrils in the presence of 50 and 100 µM diCA could
significantly reduce fibril-induced cytotoxicity to 12% and 9%. Though Aβ fibrils in the
2
5

ACCEPTED MANUSCRIPT
presence of 25 µM diCA did not reduce the cytotoxicity, our study still demonstrated that
diCA at appropriate concentration was able to rescue cytotoxicity induced by preformed
Aβ(1-42) fibrils.
2
.6 Compound 2 significantly rescues Aβ(1-42) induced cytotoxicity in transgenic C. elegans
and extends their lifespan.
We further examined the in vivo efficacy of test compounds by feeding them to
transgenic Caenorhabditis elegans that expresses Aβ(1-42) in muscle under temperature
o
regulation. When temperature shifts above 25 C, Aβ peptides start to express and accumulate
in the muscle system, leading to paralysis and death of C. elegans. Thus, the percentage of
paralysis and survival time of C. elegans can be taken as indications of the efficacy of test
compound in treatment of AD C. elegans.
o
At first, the worms were cultured and treated with individual compound at 16 C. After
2
0 h, we increased temperature to upregulate transgenic expression, which caused paralysis
and death of C. elegans within 12 h. Over 100 worms were examined with or without
paralysis in each condition, and the percentages of worms without paralysis were plotted
against time. The result showed that pretreatment with test compounds, especially diCA (2),
2
6

ACCEPTED MANUSCRIPT
significantly delayed the onset of Aβ-induced paralysis (Fig. 7). At 26 h, the survival rate of
worms treated with diCA (67.9%) was 2.5 fold increase compared with those treated in buffer
(27.6%) (Fig. 7a). The median survival time, which is a measure of 50% survival of worms,
was also analyzed, showing 25.7, 26.3 and 27.3 h in treatment with curcumin, CA and diCA,
respectively, compared with 24.2 h in buffer (control). The worms fed with diCA were able to
increase the median survival time for 3.1 h more than the buffer control. The result strongly
supported that diCA (2) is a potent compound to suppress Aβ toxicity (Fig. 7b), in agreement
with the in vitro cellular experiment. Overall, we confirm that diCA is significantly more
potent than curcumin and CA in reducing the toxicity of Aβ in vitro and in vivo. .
Fig. 7. AD-relevant C. elegans model suggests that diCA offers protection from Aβ-induced
paralysis. (a) The effect of curcumin (12), CA (13) and diCA (2) on the paralysis phenotype
observed in the transgenic C. elegans strain CL4176. The number of tested worms is about
1
00−130. (b) The median survival time was analyzed: buffer (nematode growth media), 24.2
0.4 h; curcumin, 25.7 ± 0.3 h; CA, 26.3 ± 0.3 h; diCA, 27.3 ± 0.4 h. The statistical
significance was indicated by one-way ANOVA (***, p < 0.0001; **, p < 0.005; *, p < 0.05).
±
2
7

ACCEPTED MANUSCRIPT
3
. Discussion
Caffeic acid (CA), a catechol-containing small molecule with a single ring has shown its
potential in prevention of Aβ aggregation. Other catechol-containing compounds, such as
gallic acid, dopamoin, and pyrogallol, seem to have similar effects [36]. To develop better
inhibitors for AD therapy, we synthesized a series of divalent compounds based on the
structure of CA to attain possible multivalent effect. The chemical scaffold of our designed
small molecules contains two catechol end groups that are connected with various types and
lengths of linkers. With a concern on that the catechol moiety may cause pan assay
interference (PAIN), most of our studies also included curcumin, which is devoid of catechol
moiety, as positive control.
Among the divalent compounds synthesized in this study, we found diCA (2) with a
propylenediamine linker of ~5.0 Å length efficiently inhibited Aβ fibrillization in vitro. The
inhibitory effect of divalent caffeic amide 2 (IC = 4.9 µM) is indeed better than monovalent
5
0
CA (IC = 15.6 µM) against Aβ(1-42) fibrillization.
5
0
To date, various Aβ regions have been proposed for Aβ oligomerization and aggregation.
2
8

ACCEPTED MANUSCRIPT
For example, residue 13−16 (HHQK) contributes to Aβ oligomerization and neurotoxicity,
[
(
37, 38] and binding to glycosaminoglycans [39, 40]. The hydrophobic core, residue 17−21
KLVFF) and 32−42 (IGLMVGGVVIA) are considered as crucial motifs that force Aβ to
assembly [41−43]. In addition, the flexible hinge or turn region including residue 22−31 is
also important in bringing two hydrophobic segments close to each other for formation of the
β-sheet structure. Interference with these regions could disrupt Aβ aggregation [19].
Among the dimeric caffeic amides 2−10, their Aβ inhibitory activities varied depending
on the types and lengths of linkers. Our ESI-TWIM-MS data demonstrated that diCA (2)
could bind to Aβ(1-40) and form relative stable complexes since a binomial distribution of the
bound protein peaks were observed (Fig. 5b) [34]. The MD simulation revealed that diCA
preferentially binds to the central hydrophobic region, a helix from H13 to D23 suggested
from a NMR study [44]. Our study indicated that diCA might bind to the helical region of Aβ
and disrupted the helical structure, which was visualized in one of our models (Fig. 5). The
diCA molecule might also reside in the place near the flexible hinge or turn region. It is
known the spacing between two adjacent β-stands in Aβ fibrils is 4.7 Å [2, 45, 46]. According
to our ESI-TWIM-MS experiment, Aβ dimer could form complex with diCA (2), which is
2
9