Crystal structure of thioflavin-T and its binding to amyloid fibrils: Insights at the molecular level

Article abstract of DOI:10.1039/b912396b

Combining X-ray data on thioflavin-T and theoretical calculations on its binding to a peptide model for Aβ_1-42 fibrils gives evidence of main stabilizing interactions, which influence the dihedral angle between the two moieties of thioflavin-T and thereby its fluorescence properties; these results shed new light on possible strategies for the design of dyes to bind amyloid fibrils more specifically.

Full text of DOI:10.1039/b912396b

View Article Online / Journal Homepage / Table of Contents for this issue
COMMUNICATION
www.rsc.org/chemcomm | ChemComm
Crystal structure of thioflavin-T and its binding to amyloid fibrils:
insights at the molecular levelw
Cristina Rodrı
´
Angel Alvarez-Larena, Hugo Gutierrez-de-Teran, Mariona Sodupe* and
´
Pilar Gonzalez-Duarte*
guez-Rodrı
´
c
guez,a Albert Rimola,b Luis Rodrı
´
guez-Santiago,a Piero Ugliengo,b
d
a
´
´
´
´
a
Received (in Cambridge, UK) 26th June 2009, Accepted 2nd December 2009
First published as an Advance Article on the web 4th January 2010
DOI: 10.1039/b912396b
Combining X-ray data on thioflavin-T and theoretical calculations
on its binding to a peptide model for Ab1–42 fibrils gives evidence
of main stabilizing interactions, which influence the dihedral
angle between the two moieties of thioflavin-T and thereby its
fluorescence properties; these results shed new light on possible
strategies for the design of dyes to bind amyloid fibrils more
specifically.
The presence of amyloid fibrils is a key neuropathological
feature in more than 20 diseases, including Alzheimer’s disease
(AD).1 Most of these protein aggregates have a generic
structure that consists of a continuous b-sheet.2 The fluores-
cent dye thioflavin-T (ThT) has been commonly used to stain
amyloid deposits in tissues3 and to characterise the presence of
amyloid fibrils.4 Chemically, ThT is the chloride salt of a cation
(ThT+) that consists of an N-methylated benzothiazole frag-
ment linked to a dimethylaniline ring (Fig. 1). The dihedral
angle between these two planar moieties, f, largely determines
the spectral properties of this dye, as shown by Stsiapura et al.5
from a computational study of the torsional relaxation of ThT+
in the excited state. However, the mechanism of fluorescence
enhancement upon ThT+ binding to amyloid fibrils and the
molecular understanding of this binding are still objects of
intensive study.6 Both issues are critical in order to design
new dyes to bind amyloid fibrils more specifically,7,8 and for
the development of diagnostic and therapeutic agents.
Fig. 1 Unit cells of 1 (a) and 2 (b): I4 and IÀ anions in purple;
2À
chlorine, nitrogen and sulfur atoms in green, blue and yellow, respectively.
shed light onto ThT+-binding interactions, we have under-
taken an experimental and theoretical study of the structure of
unbound ThT+. Subsequently, we have examined its binding
to the fibrillar core structure of Ab1–42 peptide11 by combining
docking and molecular dynamics (MD) strategies with
quantum chemical calculations.
Recently, we have shown that incorporation of chelating
properties into well-known intercalation compounds such as
ThT+ may be a successful strategy for developing the next
generation of agents for AD theragnosis.9 Remarkably, there
are no X-ray structural data on the free ThT+ but only when
bound to the Torpedo californica acetylcholinesterase
(TcAChE) protein,10 one of the few examples showing that
ThT+ also binds to non-b-sheet cavities. Overall, in order to
Treatment of (ThT)Cl with NaI, as detailed in the
experimental section (see ESIw), yielded X-ray quality crystals
of (ThT)2I4Á2CHCl3 (1) and (ThT)IÁCHCl3 (2).12 The unit cells
are shown in Fig. 1. Both I4 (1) and IÀ (2) anions establish
2À
weak C–HÁ Á ÁI interactions with the CHCl3 solvent molecules.
I42À, whose formation is probably due to aerobic oxidation,
a
´
Departament de Quımica, Universitat Autonoma de Barcelona,
08193 Bellaterra, Barcelona, Spain.
`
shows
a
centrosymmetric essentially linear geometry,
consistent with reported data.13,14 This anion appears to be
essential for the effective packing arrangement of the ThT+
cations in 1, which is formed in much shorter time than 2. Our
unsuccessful attempts to obtain any crystal structure under
other experimental conditions evidence the relevance of the
iodide anions and CHCl3 solvent in the formation of X-ray
quality crystals. Remarkably, the closest related analogs of
ThT+ with known crystal structures are not cationic species
but neutral molecules.15–17
E-mail: mariona.sodupe@uab.cat, pilar.gonzalez.duarte@uab.cat
b Dipartimento di Chimica IFM and NIS Centre of Excellence,
`
Universita degli Studi di Torino, 10125 Torino, Italy
Servei de Difraccio de Raigs X, Universitat Autonoma de Barcelona,
c
´
`
Barcelona, Spain
d
´
Fundacion Publica Galega de Medicina Xenomica, CHUS,
15786 Santiago de Compostela, Spain
´
´
w Electronic supplementary information (ESI) available: Experimental,
X-ray diffraction and computational details. CCDC 728298 and
728299. For ESI and crystallographic data in CIF or other electronic
format see DOI: 10.1039/b912396b
ꢀc
This journal is The Royal Society of Chemistry 2010
1156 | Chem. Commun., 2010, 46, 1156–1158

View Article Online
in the crystal structures. Within this context, a recent study on
the X-ray structure of ThT+ bound to TcAChE10 shows the
ThT+ ring system in a coplanar position packing parallel to
aromatic residues. This packing causes a decrease in torsional
relaxation of ThT+ and, hence, an increase in its fluorescence.
Despite the wide use of ThT+ as a fluorescent dye to
determine the presence of amyloid fibrils, studies on the corres-
ponding binding mode at the molecular level are scarce.21,22 In
order to make a further step, we have examined the interaction
of ThT+ with the fibrillar core structure of Ab1–42 peptide, as
deduced from NMR data.11 Thus, we have performed a series
of automated docking, molecular dynamics and quantum
mechanical calculations. The automated blind docking explora-
tion involved 100 docking runs on each one of the ten protein
conformers deposited under the PDB code 2BEG. In all 1000
solutions the ThT+ cation was placed parallel to the long axis of
the fibrils, in agreement with anisotropic fluorescence emission
studies.21 After clustering all docking possibilities, we identified
six probable binding modes, which are located on four different
binding sites, two within the inner core of the fibrils and the
other two on the external surface (Fig. S1, ESIw). A deeper
exploration of each binding mode by MD and binding free
energy calculations, using the linear interaction energy
approach23 (see ESI for detailsw), shows that the two external
binding sites, in which ThT+ is accommodated into the crevices
defined either by residues 33–35 (pose A) or 35–37 (pose B1), are
the most favorable (Fig. 3). Remarkably, in both locations the
planar dimethylamino moiety of ThT+ is positioned nearly
parallel to the main backbone of the polypeptide chain. The
ease of rotation around the C–C bond of ThT+ accounts for the
orientation of the benzothiazole fragment, with the NMe+
group pointing away from the crevice in pose A and towards
the Met side chain in pose B1.
Fig. 2 Front and top view of the stacking arrangements in 1 (a) and 2
(b). X-Ray distances (in A) between centroids are indicated by dashed
lines. Periodic B3LYP values are in brackets.
The only crystallographically independent ThT+ cation in 1
and 2 show very similar geometric parameters, the dihedral
angle f between the two planar benzothiazole and dimethyl-
aniline moieties being 23.8(2)1 in 1 and 20.0(2)1 in 2. In both
crystal structures the ThT+ cations are packed in columnar
stacks, neighbouring cations having a head to tail disposition
and their positive charge being on opposite sides to minimize
Coulombic repulsion. Cations defining a stack in 1 are related
by a twofold screw axis, while in 2 neighbouring cations are
related by an inversion centre (Fig. 2). In both crystal structures
the distances between stacked cations and their p–p facing
reveals a trade-off between ring electric multipole and dispersion
interactions. In 1 the main overlap is observed between thiazolyl
and dimethylaniline rings (Th–Ph), whereas in 2 two different
overlaps are present: a Th–Th interaction, and a partial Bz–Ph
overlap (Bz denotes the benzene ring of benzothiazole).
The values of f in 1 (241) and 2, (201), are smaller than that
previously computed for the isolated ThT+ cation,5 which
ranges from 321 to 451, depending on the adopted level of
theory. This suggests that packing effects, particularly stacking
interactions between ThT+ cations in the crystal, are
responsible for the decrease of this dihedral angle f which,
as mentioned, largely determines the spectral properties of this
dye. To investigate this point, we have performed B3LYP
calculations18,19 of the ThT+ cation either isolated or in
crystals 1 and 2. Periodic geometry optimizations were
performed only for the internal atomic positions, the lattice
parameters being frozen at the experimental values. This
procedure has been shown to be a cost effective strategy for
describing molecular crystals where dispersion forces are
relevant20 (see detailed discussion in ESIw) and it also holds
for the present systems, since main optimized B3LYP para-
meters are in good agreement with the X-ray data, the
computed distances differing by 0.05 A at the most (Fig. 2).
The f value we have calculated for the isolated ThT+ (351)
agrees with previously reported data5 and is substantially
larger than that we have obtained for 1 (201) and 2 (191).
Noticeably, the latter values are consistent with those obtained
from X-ray data, 241 (1) and 201 (2), which validates the
reliability of our calculations and shows the importance of
packing effects. Overall, the intrinsic ease of rotation around
the C–C bond, which at the B3LYP level is computed to be
5 kcal molÀ1, and the relatively large stacking interactions in
the crystals account for the increase in planarity of ThT+
Among the two positions, pose B1 is more stable by
2.6 kcal molÀ1 and encloses a ThT+ cation with increased
planarity (441) if compared with unbound ThT+ (651) as
simulated under the same conditions. This latter f value
resulting from the adopted OPLS force field is significantly
higher than that calculated with B3LYP for ThT+ in the gas
phase (351). Therefore, to keep coherence with the quantum
mechanical method used, we have performed optimizations
at the B3LYP-D level for ThT+ interacting with a model
of a b-sheet structure, made by five strands of the
CH3(NHCOCH2)3NHCOCH3 peptide, which mimics the
interaction with the polypeptide chain. Results confirm that
Fig. 3 External binding sites of ThT+ bound to Ab1–42 from
molecular dynamics calculations. ThT+ is located in the crevices
defined by the residues 33–35 (pose A) or 35–37 (pose B1).
ꢀc
This journal is The Royal Society of Chemistry 2010
Chem. Commun., 2010, 46, 1156–1158 | 1157

View Article Online
the two moieties: (a) the p–p interaction between the dimethyl-
aniline moiety and the peptide backbone, which can be present
regardless of the protein originating the b-sheet structure, and (b)
the CH–p interaction between the Gly residue, the only one
present in the model peptide, and the benzothiazole moiety. This
second interaction, which influences the dihedral angle between
the two moieties of ThT+ and thus its fluorescence properties
due to a decrease in torsional relaxation, is dependent on the
amino acid residues of the peptide chain. We believe that
modulation of this latter interaction by proper molecular design
may bring the discovery of new dyes for more specific binding to
amyloid fibrils as well as to new diagnostic agents.
We thank the MICINN (CTQ2008-06381/BQU) and DURSI
(SGR2009-68 and SGR2009-638) for financial support.
Fig. 4 B3LYP-D optimized structures of ThT+ interacting with a
model of the b-sheet structure. Distances are in A. Front views show
the distance between the centroid of dimethylaniline and the plane
defined by the peptide bond atoms immediately below. Side views
show the computed distances between the hydrogen of the Gly residue
and the ring centroids of benzothiazole.
Notes and references
1 H. J. R. Jakob-Roetne and H. Jacobsen, Angew. Chem., Int. Ed.,
2009, 48, 3030–3059.
2 F. Chiti and C. M. Dobson, Annu. Rev. Biochem., 2006, 75, 333–366.
3 P. S. Vassar and C. F. Culling, Arch. Pathol., 1959, 68, 487–498.
4 H. LeVine, 3rd, Protein Sci., 1993, 2, 404–410.
5 V. I. Stsiapura, A. A. Maskevich, V. A. Kuzmitsky, K. K. Turoverov
and I. M. Kuznetsova, J. Phys. Chem. A, 2007, 111, 4829–4835.
6 A. Hawe, M. Sutter and W. Jiskoot, Pharmacol. Res., 2008, 25,
1487–1499.
7 A. Nordberg, Lancet Neurol., 2004, 3, 519–527.
8 R. Leuma Yona, S. Mazeres, P. Faller and E. Gras, ChemMedChem,
2008, 3, 63–66.
the dimethylaniline ring remains parallel to the peptide
backbone, the distance between the centroid of the benzene ring
and the mean plane defined by the atoms involved in the closest
peptide bond being 3.06 A (Fig. 4a). The concomitant decrease of
f down to 211 is mainly due to the stabilizing interaction between
the benzothiazole fragment and the closest H(Gly), the distance
between the centroid of the benzene ring and hydrogen being
2.94 A. Overall, the minimum energy structure is mainly deter-
mined by p-stacking and weak CH–p interactions. Indeed, if
dispersion forces are not taken into account in the calculation by
adopting the plain B3LYP, the optimized structure shows that
dimethylaniline is no longer parallel to the peptide bond, the
distance between the centroid of benzene and the peptide bond
plane increasing up to 4.0 A and f increasing to 321. Moreover,
we have located a second nearly degenerate energy minimum
(0.2 kcal molÀ1 below the previous one) with f being approxi-
mately 1401 (Fig. 4b). In this case, the dimethylaniline ring
remains parallel to the peptide backbone and CH–p interactions
are maintained with the nearest H(Gly) residue. Remarkably, the
energy cost of internal rotation of ThT+ bound to the fibril,
estimated by performing a constrained geometry optimization
with the dihedral angle of ThT+ fixed at 901, is 7 kcal molÀ1. This
value, although somewhat larger than that obtained for unbound
ThT+ (5 kcal molÀ1) is not large enough to prevent internal
rotation. These results suggest that interactions between benzo-
thiazole and the side chains of the fibril are the ones responsible
for the decrease of torsional relaxation in bound ThT+.
In summary, the dihedral angle f of ThT+ in crystals 1 (241)
and 2 (201) can be explained by considering p-stacking inter-
actions between adjacent cations. This angle is very close to that
computed for ThT+ bound to a model of Ab1–42 fibrils, where
p–p and CH–p interactions account for the localization and
orientation of ThT+. These features are fully consistent with
anisotropic fluorescence emission studies showing that ThT+
binds to the fibrils with its long axis parallel to that of fibrils.21
The computational results give a further insight into this binding
at a molecular level and show the relevance of dispersion
forces which, as recently reviewed,24 play a fundamental role in
biological sciences. Concerning ThT+ binding to fibrils, two
main stabilizing interactions are present, each involving one of
9 C. Rodrı
´
guez-Rodrı
A. Alvarez-Larena, V. Lloveras, J. Vidal-Gancedo, S. Ventura,
J. Vendrell, M. Sodupe and P. Gonzalez-Duarte, J. Am. Chem.
Soc., 2009, 131, 1436–1451.
guez, N. Sanchez de Groot, A. Rimola,
´ ´
´
´
10 M. Harel, L. K. Sonoda, I. Silman, J. L. Sussman and
T. L. Rosenberry, J. Am. Chem. Soc., 2008, 130, 7856–7861.
11 T. Luhrs, C. Ritter, M. Adrian, D. Riek-Loher, B. Bohrmann,
H. Dobeli, D. Schubert and R. Riek, Proc. Natl. Acad. Sci. U. S. A.,
2005, 102, 17342–17347.
12 Crystal data. For 1: C36H40Cl6I4N4S2, monoclinic, a = 36.297(3),
b = 7.1666(5), c = 19.5063(13) A, b = 112.525(1)1, V = 4687.0(6) A3,
T = 296 K, space group C2/c (no. 15), Z = 4, 15 766 reflections
measured, 5775 unique (Rint = 0.035) which were used in all calcula-
tions, R(F) = 0.052 for 3500 reflections with I 4 2s(I), Rw(F2) =
0.143 for all data. For 2: C18H20Cl3IN2S, M = 529.67, monoclinic,
a = 7.4500(5), b = 14.7704(9), c = 19.5483(12) A, b = 94.431(1)1,
V = 2144.7(2) A3, T = 296 K, space group P21/n (no. 14), Z = 4,
14 563 reflections measured, 5255 unique (Rint = 0.036) which were
used in all calculations, R(F) = 0.042 for 3608 reflections with
I 4 2s(I), Rw(F2) = 0.096 for all data.
13 C. A. Ilioudis and J. W. Steed, CrystEngComm, 2004, 6, 239–242.
14 D. L. Long, H. M. Hu, J. T. Chen and J. S. Huang, Acta
Crystallogr., Sect. C: Cryst. Struct. Commun., 1999, 55, 339–341.
15 X.-J. Chen, P.-R. Yu, W.-S. Wang and B.-L. Liu, Acta Crystallogr.,
Sect. E: Struct. Rep. Online, 2007, 63, o595–o596.
16 W.-S. a. G. Wang, Min, Chen, Xiangji, Liu and Boli, Acta
Crystallogr., Sect. E: Struct. Rep. Online, 2006, 62, o5668–o5669.
17 Y. Zhang, Z.-H. Su, Q.-Z. Wang and L. Teng, Acta Crystallogr.,
Sect. E: Struct. Rep. Online, 2008, 64, o2065.
18 C. Lee, W. Yang and R. G. Parr, Phys. Rev. B: Condens. Matter,
1988, 37, 785–789.
19 A. D. Becke, J. Chem. Phys., 1993, 98, 5648–5652.
20 B. Civalleri, C. M. Zicovich-Wilson, L. Valenzano and
P. Ugliengo, CrystEngComm, 2008, 10, 405–410.
21 M. R. Krebs, E. H. Bromley and A. M. Donald, J. Struct. Biol.,
2005, 149, 30–37.
22 C. Wu, Z. Wang, H. Lei, Y. Duan, M. T. Bowers and J. E. Shea,
J. Mol. Biol., 2008, 384, 718–729.
23 J. Aqvist, C. Medina and J. E. Samuelsson, Protein Eng., Des. Sel.,
1994, 7, 385–391.
24 Special Issue—Stacking interactions, Phys. Chem. Chem. Phys, ed.
P. Hobza, 2008, 10(19), 2561–2868.
ꢀc
This journal is The Royal Society of Chemistry 2010
1158 | Chem. Commun., 2010, 46, 1156–1158