In vivo optical imaging of amyloid aggregates in brain: Design of fluorescent markers

Article abstract of DOI:10.1002/anie.200500845

(Figure Presented) Routine diagnostics and studies of Alzheimer's disease might benefit form the noninvasive optical imaging of amyloid-β plaques in the brain. A rational design strategy for in vivo amyloid-imaging agents that enter the brain and selectiv

Full text of DOI:10.1002/anie.200500845

Communications
Fluorescent Probes
agents. These contrast agents make use of the inverse
fourth-power relationship between the wavelength and light
scattering, thus allowing the longer-wavelength light pene-
tration through the living tissues necessary for direct brain
imaging.[4a] Long-wavelength detection methods also benefit
from the low auto-fluorescence of biological matter beyond
600 nm.[4] The requirements for a successful optical marker of
AD are: 1)a suitable wavelength interval of absorption and
emission (600–800 nm), 2) the ability to rapidly enter the
brain after intravenous injection, and 3)specific labeling of
the amyloid-b deposits with rapid clearing of the unbound
dye. Another unique advantage of optical imaging is the
possibility to attain substantial differences in the photo-
physical/optical properties between the bound and unbound
forms of the marker (requirement 4). This would allow a
significant increase in imaging contrast which cannot be
exploited with techniques based on radioligands.
DOI: 10.1002/anie.200500845
In Vivo Optical Imaging of Amyloid Aggregates
in Brain: Design of Fluorescent Markers**
Evgueni E. Nesterov, Jesse Skoch, Bradley T. Hyman,
William E. Klunk, Brian J. Bacskai, and
Timothy M. Swager*
Neuroimaging is emerging as a way to noninvasively identify
and monitor neurodegenerative diseases during the preclin-
ical and early clinical stages.[1] Alzheimerꢀs disease (AD)
stands out among the neurodegenerative diseases as the
fourth leading cause of death in the United States and the
most common cause of acquired dementia.[2] Early detection
of AD is imperative in enabling the understanding and
clinical treatment of this disorder, as well as in preventing its
progression. The characteristic signature of AD is the
deposition of amyloid-b plaques and neurofibrillary tangles
in the patientꢀs brain,[3] which parallel the disease but can only
be diagnosed with certainty after death by an autopsy.
Among the known amyloid-staining compounds, Congo
Red (CR)provides historically the most standardized way of
A significant advance in the field has been the develop-
ment of radiolabeled small-molecule agents capable of enter-
ing the brain and specifically targeting plaques and tangles for
imaging with positron emission tomography (PET)and
single-photon
emission
computerized
tomography
(SPECT).[1] A major limitation of these methods is the
requirement for the markers to be labeled with short-lived
isotopes, such as 11C with a half-life of about 20 min for PET.
An attractive noninvasive alternative is in vivo optical imag-
ing using specific far-red (near-IR)fluorescent contrast
[*] Dr. E. E. Nesterov,[+] Prof. Dr. T. M. Swager
Department of Chemistry and
Institute for Soldier Nanotechnologies
Massachusetts Institute of Technology
Cambridge, MA 02139 (USA)
Fax: (+1)617-324-0505
E-mail: tswager@mit.edu
J. Skoch, Prof. Dr. B. T. Hyman, Dr. B. J. Bacskai
Department of Neurology/Alzheimer’s Disease
Research Laboratory
Massachusetts General Hospital
staining amyloid plaques, and is still employed in post mortem
histological analysis of AD brain, as the binding is specific.[5]
Thioflavin T (ThT)is another dye to use in analysis of
aggregated amyloid proteins. It binds slightly weaker than
CR, but makes up for this deficiency by exhibiting a green
fluorescence, that becomes more than 1000 times brighter
upon binding to amyloid plaques.[6] The understanding of
what makes these simple molecules so specific to senile
plaques is the starting point for the rational design of
improved markers. Studies[7] suggest that the origin of the
specific binding of CR to amyloid-b aggregates is due to the
combination of electrostatic interactions between the nega-
tively charged CRꢀs sulfonate groups with the positively
charged amino acid residues in the antiparallel protein
114 16th Street, Charlestown, MA 02129 (USA)
Prof. Dr. W. E. Klunk
Department of Psychiatry
University of Pittsburgh School of Medicine
Pittsburgh, PA 15213 (USA)
[+] Current address:
Department of Chemistry
Louisiana State University
Baton Rouge, LA 70803 (USA)
[**] This research was supported by NASA and the U.S. Army through
the Institute for Soldier Nanotechnologies, under contract DAAD-
19-02-D-0002 with the U.S. Army Research Office (TMS), by NIH
through grant EB00768 (BJB).
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
5452
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 5452 –5456

Angewandte
Chemie
b sheets, and CRꢀs shape is complementary to features in an
amyloid fibrilꢀs surface. The latter recognition likely involves
hydrophobic interactions with the planarized aromatic p
system of the dye molecule.[7c] Although the binding of ThT to
amyloids has received less attention, it is likely to involve
similar shape-specific recognition.[6,8] Thus, it is our opinion
that the presence of hydrophobic planarized (or easily
planarizable) p system is an important design feature to
obtain high binding specificity.
The highly polarizable bridge was chosen to provide a
substantial bathochromic shift in both absorption and emis-
sion spectra of the dye. The rigid rodlike aromatic core of
NIAD-4 is hydrophobic and readily planarizable, and has a
shape similar to that of ThT for achieving a high binding
specificity to amyloid aggregates. In unbound NIAD-4 free
rotation around the single bonds leads to facile vibrationally
coupled radiationless excited-state deactivation. Reduction of
these processes upon binding to the protein aggregate was
expected to enhance its fluorescence quantum yield. The
donor and acceptor groups were chosen to balance between
the need for a high electron-donating/accepting ability to
achieve a substantial spectral bathochromic shift and the
requirement to avoid charged or easily ionizable groups to
improve the BBB permeability. The dicyanomethylene group
was chosen as the acceptor, and the hydrophilic p-hydroxy-
phenyl group was selected as a donor to compensate for the
high hydrophobicity of the bridge. Both groups are relatively
small and were instrumental in maintaining the low molecular
weight of NIAD-4.
A potential biomarker must be able to rapidly enter the
brain after an intravenous injection. To exhibit high blood–
brain barrier (BBB)permeability, the molecule needs to be
relatively lipophilic (but not highly lipophilic as this will lead
to its aggregation and accumulation in blood proteins and red
blood cell membranes), and should not bear ionic groups.
High BBB permeability also requires the marker to be
relatively small, with an upper molecular weight limit of about
600 Da. These simple empirical rules are extremely robust,
and, for example, both CR and ThT, as charged molecules, are
not capable of crossing the BBB. It is also highly probable that
any charged molecules like compound 2C40, a recently
proposed fluorescent marker for AD found by combinatorial
screening of a library of 320 compounds,[9] will be similarly
incapable of in vivo imaging. In contrast, substantial increase
of the BBB permeability was achieved with uncharged
analogues of CR and ThT. For example, in the CR analogue
Chrysamine G (CG), the replacement of the sulfonate group
by a less ionizable carboxy group significantly increased the
BBB permeability without compromising the high binding
specifity.[10] A ThT uncharged analogue, 6-OH-BTA-1, which
is commonly referred to as Pittsburgh compound B (PIB), is
highly efficient both in crossing the BBB and in selective
binding to AD amyloid aggregates, and is currently under
clinical tests as an amyloid marker with PET imaging.[11]
To develop an effective in vivo imaging dye, we have
merged these empirical requirements with electronic struc-
ture design to produce a compound possessing a suitable
spectral range of the absorption and emission bands along
with a potential to produce additional imaging contrast. We
have employed a classical push–pull architecture with termi-
nal donor and acceptor moieties that are interconnected by a
highly polarizable bridge. Various donor and acceptor groups
can be used to manipulate the relative energies of HOMO
and LUMO, with better donor and better acceptor leading to
a smaller HOMO–LUMO gap and to the desired long
wavelength absorption/emission bands. Finally, to obtain
greater imaging contrast, we required a molecule to possess
The synthesis of NIAD-4 is shown in Scheme 1. The silyl-
protected aldehyde 4 was prepared by Stille coupling.
Scheme 1. Synthesis of dye NIAD-4: a) [PdCl2(PPh3)2] (3 mol%), tolu-
ene, 708C, 36 h; b) nBuLi, TMEDA, hexanes, ꢁ708C, 2 h; c) DMF,
ꢁ708C!RT, 12 h; d) CH2(CN)2, EtOH, piperidine (cat.), 608C, 18 h;
e) HCl, 608C, 6 h. DMF: N,N-dimethylformamide; TMEDA: N,N,N’,N’-
tetramethylethylenediamine.
a
substantial degree of conformational freedom when
unbound in solution, but it would be conformationally
restricted upon binding to the protein. Such a “rigidification”
decreases the vibrational-rotational processes that couple the
excited and ground states, thereby decreasing the radiation-
less decay rate and increasing the fluorescence quantum yield
of the protein-bound molecule.[12]
The proof-of-approach near-IR Alzheimerꢀs dye NIAD-
4[13] was designed to satisfy all four of the requirements above.
This dye with a molecular weight of 334 Da is likely to be
among the simplest possible structures. It uses a dithienyl-
ethenyl p-conjugated bridge between donor and acceptor.
Condensation with malononitrile followed by in situ depro-
tection afforded NIAD-4 as a dark-red material, which
exhibits a bright red fluorescence in methanol solution
(Figure 1). This contrast marker is slightly soluble in aqueous
media, although it forms micelles/aggregates at the concen-
trations above 10 mm.[14] The aqueous solutions of NIAD-4
only have trace emission (quantum yield ꢀ 0.00008), which
may be in part due to the greater stabilization in aqueous
media of the non-emissive intramolecular charge transfer
state as opposed to the emissive planar excited state.[15] An
in vitro spectral study with aggregated synthetic amyloid-b 1–
Angew. Chem. Int. Ed. 2005, 44, 5452 –5456
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
5453

Communications
that of high-affinity amyloid-binding compounds like PIB
(Ki = 4.3 nm).[11b] The excellent performance of NIAD-4 for
in vivo staining of amyloid plaques was demonstrated by
using transgenic mice with AD-like pathology prepared with
cranial windows to allow direct monitoring of the brain
surface.[16] The dye readily crosses the BBB after an intra-
venous injection and labels specifically both the plaques and
cerebrovascular amyloid angiopathy in the living brain thus
making them easily detectable by the characteristic red
fluorescence (see Figure 3 and Supporting Information).
The high specificity of labeling the amyloid deposits is also
clearly observed with in situ histochemical staining of fixed
sections from transgenic mouse brain (Figure 3). In both
in vivo and in situ experiments, the fluorescent images reveal
exact position and size of the aggregated amyloid deposits,
therefore providing the best demonstration of the high
labeling specificity of NIAD-4.
Figure 1. Absorption and fluorescence spectra of NIAD-4 in MeOH so-
lution. (Extinction coefficient e(475 nm) 35700; fluorescence quantum
yield 0.15.)
40 protein displayed a significant red shift (Dl ꢀ 70 nm)of the
absorption maximum along with about 1.3-fold increase in its
extinction coefficient (Figure 2). This phenomenon is a
Figure 3. In vivo targeting and in situ specificity of NIAD-4 for amyloid
deposits in the brain. a) A fluorescent image of a coronal section of
brain from an aged APP transgenic mouse labeled with a 10 mm solu-
tion of NIAD-4 in DMSO/propylene glycol for 15 min at RT. b) In vivo
fluorescent detection of both senile plaques and amyloid angiopathy
using multiphoton microscopy immediately following intravenous
injection of 2 mgkgꢁ1 of NIAD-4. The images demonstrate the egress
of dye through the BBB in a living mouse as well as the specific, sensi-
tive targeting of amyloid-b deposits in the brain.
Figure 2. Absorption and fluorescence spectra of NIAD-4 in aqueous
PBS solution in the presence of aggregated amyloid-b protein (solid
line) and without the protein (dashed line). Concentrations: NIAD-4
4.1 mm, amyloid-b ꢀ10 mm. PBS: phosphate buffered saline.
The remarkably strong bathochromic shift in absorption
concomitant with the increase of absorption intensity upon
binding combine for the high contrast optical imaging. To
support our model that these changes are in part due to the
planarization of NIAD-4 upon binding, we have calculated
the geometry-optimized structure of the dye molecule
(Figure 4a). Although both thienyl rings are nearly coplanar
(dihedral angle 4.58), the benzene ring is markedly twisted
with respect to the plane of the neighboring thienyl group
(dihedral angle 24.58). This geometry still allows a p conju-
gation, and both HOMO and LUMO are substantially
delocalized over the entire molecule (Figure 4c). Interest-
ingly, just a small geometry change towards further planari-
zation (which is likely to happen upon binding to amyloid
fibrils)significantly increases the conjugation, resulting in
even more delocalized frontier orbitals and a smaller
HOMO–LUMO gap. The computed electrostatic potential
distribution over the electron density surface (Figure 4b)is
very similar to that of CR, with the highly hydrophobic
neutral core and negative charge density localized at the
consequence of the strong binding between the dye and the
aggregated amyloid fibrils. The fluorescence maxima of the
bound and unbound forms do not differ appreciably, and, as
expected, the binding is accompanied by a dramatic enhance-
ment of the fluorescent emission (about 400-fold, quantum
yield 0.05), which is visible by simple visual inspection even at
the micromolar concentrations used in the experiment.
Apparently, the protein-bound dye is preorganized in a
geometry that is similar to the planarized geometry of the
emissive excited state after it has reached its equilibrium
nuclear coordinates. Furthermore, binding to the amyloid
fibrils places the dye in a lower dielectric constant environ-
ment that decreases detrimental relaxation into a non-
emissive intramolecular charge transfer state.[15]
The binding studies with artificially aggregated amyloid
protein assays revealed that NIAD-4 binds to the same site as
BTA-1 with a binding affinity Ki of 10 nm. This affinity is
much higher than that of ThT (Ki = 580 nm), and is close to
5454
www.angewandte.org
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 5452 –5456

Angewandte
Chemie
Keywords: Alzheimer’s disease · dyes/
.
pigments · fluorescent probes · imaging agents ·
synthesis design
[1] S. T. DeKosky, K. Marek, Science 2003, 302,
830 – 834.
[2] “Alzheimer Disease”: P. Nowotny, J. M. Kwon,
A. M. Goate in Encyclopedia of Life Sciences,
Macmillan Publishers, Nature Publishing
Group, 2002.
[3] P. M. Gorman, A. Chakrabartty, Biopolymers
2001, 60, 381 – 394.
[4] a)“Near-Infrared Dyes for High Technology
Applications”: E. Terpetschnig, O. S. Wolfbeis,
NATO ASI Ser. Ser. 3 1998, 52, 161 – 182; b)A.
Becker, C. Hessenius, K. Licha, B. Ebert, U.
Sukowski, W. Semmler, B. Wiedenmann, C.
Grötzinger, Nat. Biotechnol. 2001, 19, 327 – 331.
[5] G. T. Westermark, K. H. Johnson, P. Wester-
mark, Methods Enzymol. 1999, 309, 3 – 25.
[6] E. S. Voropai, M. P. Samtsov, K. N. Kaplevskii,
A. A. Maskevich, V. I. Stepuro, O. I. Povarova,
I. M. Kuznetsova, K. K. Turoverov, A. L. Fink,
V. N. Uverskii, J. Appl. Spectrosc. 2003, 70,
868 – 874.
Figure 4. Optimized ground-state geometry of NIAD-4 (a), color-coded electrostatic
potential mapped over electron density surface (b), and frontier molecular orbitals (c):
HOMO (left) and LUMO (right). The computations were performed in gas phase at
B3LYP/6-31G(df,p) level of theory. For this geometry, the computed absorption maxi-
mum (TD-B3LYP/6-31G(df,p)) was found to be 481 nm, which is in excellent agreement
with the experimental value (475 nm).
[7] a)I. Roterman, M. Krol, M. Nowak, L.
Konieczny, J. Rybarska, B. Stopa, B. Piekarska,
G. Zemanek, Med. Sci. Monit. 2001, 7, 771 –
784; b)J. Sajid, A. Elhaddaoui, S. Turrell, J.
Mol. Struct. 1997, 408/409, 181 – 184; c)R.
moleculeꢀs termini. This feature may be in part responsible for
Khurana, V. N. Uversky, L. Nielsen, A. L. Fink, J. Biol. Chem.
2001, 276, 22715 – 22721; d)D. B. Carter, K.-C. Chou, Neuro-
biol. Aging 1998, 19, 37 – 40; e)W. E. Klunk, J. W. Pettegrew,
D. J. Abraham, J. Histochem. Cytochem. 1989, 37, 1273 – 1281.
[8] H. LeVine III, Amyloid 1995, 2, 1 – 6.
[9] Q. Li, J.-S. Lee, C. Ha, C. B. Park, G. Yang, W. B. Gan, Y.-T.
Chang, Angew. Chem. 2004, 116, 6491 – 6495; Angew. Chem. Int.
Ed. 2004, 43, 6331 – 6335.
the high binding selectivity of NIAD-4 to aggregated amyloid
protein.[17]
In conclusion, we have demonstrated that the targeted
rational design of in vivo fluorescent markers for amyloid
protein aggregates can be successfully accomplished with a set
of very simple qualitative rules. The designed contrast agent
NIAD-4 shows remarkable promise for the optical imaging of
amyloid plaques in brain. It may also be useful for the
detection of peripheral amyloidoses. Clearly, its practical
application in human Alzheimerꢀs disease research and
diagnostics requires further red-shifting of its spectral char-
acteristics and increasing fluorescence quantum yield to allow
truly noninvasive monitoring of the brain surface, without the
need for the cranial window. It also requires detailed and
thorough toxicity studies which we have not performed at this
stage.[18] Still this is a remarkable milestone demonstrating
great power of the simple design principles outlined herein for
development of a practical AD biomarker. Regardless of the
spectral properties of this new compound, a radiolabeled
version of NIAD-4 may also be advantageous for PET or
SPECT imaging. Directed modifications of the structure of
NIAD-4 with the purpose to further red-shift the spectral
characteristics, increase fluorescence quantum yields, and
generally improve performance, as well as gain deeper
understanding of these markers, are currently underway.
[10] W. E. Klunk, M. L. Debnath, J. W. Pettegrew, Neurobiol. Aging
1995, 16, 541 – 548.
[11] a)W. E. Klunk, Y. Wang, G.-F. Huang, M. L. Debnath, D. P.
Holt, C. A. Mattis, Life Sci. 2001, 69, 1471 – 1484; b)C. A.
Mattis, Y. Wang, D. P. Holt, G.-F. Huang, M. L. Debnath, W. E.
Klunk, J. Med. Chem. 2003, 46, 2740 – 2754; c)W. E. Klunk, H.
Engler, A. Nordberg, Y. Wang, G. Blomqvist, D. P. Holt, M.
Bergström, I. Savitcheva, G.-F. Huang, S. Estrada, B. Ausꢁn,
M. L. Debnath, J. Barletta, J. C. Price, J. Sandell, B. J. Lopresti,
A. Wall, P. Koivisto, G. Antoni, C. A. Mathis, B. Lꢂngström,
Ann. Neurol. 2004, 55, 306 – 319.
[12] The phenomenon of increasing the emission quantum yield in
viscous solvents, where a conformationally labile molecule is
vibrationally-restricted, is well documented for ThT (ref. [6]), as
well as for some other fluorescent molecules (K. A. Willets, O.
Ostroverkhova, M. He, R. J. Twieg, W. E. Moerner, J. Am.
Chem. Soc. 2003, 125, 1174 – 1175).
[13] NIAD-4 is one of a series of contrast agents that we have been
developing and testing. Additional NIAD compounds will be
reported in the future.
[14] The 10 mm aggregation threshold, determined by a simple visual
inspection, is a very approximate estimate, and strongly depends
on the way the solution was prepared. The solutions used in the
experiments reported herein had at least twice lower concen-
tration, and their absorbance did not change after passing
through a syringe filter with 0.2 mm pore size. Based on this, we
can reasonably assume that NIAD-4 was not aggregated in the
solutions.
Received: March 7, 2005
Revised: July 16, 2005
Published online: August 1, 2005
Angew. Chem. Int. Ed. 2005, 44, 5452 –5456
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
5455

Communications
[15] We are currently conducting computational studies to gain
further insight into this phenomenon.
[16] B. J. Bacskai, G. A. Hickey, J. Skoch, S. T. Kajdasz, Y. Wang, G.-
F. Huang, C. A. Mathis, W. E. Klunk, B. T. Hyman, Proc. Natl.
Acad. Sci. USA 2003, 100, 12462 – 12467.
[17] Alternatively, the spectral changes upon binding to amyloid may
be explained by change in protonation of the phenolic hydroxy
group upon binding of the dye to a protein aggregate. This
behavior should be similar to the deprotonation at higher pH
values, which can be employed to study the deprotonation effect.
Although these experiments showed an anticipated change in
optical properties at increasing pH (Figure S1 in the Supporting
Information), the overall trend does not correspond to the
spectral changes found upon binding to amyloid. This allowed us
to safely rule out the explanation based on the protonation
change.
[18] Injection of NIAD-4 into living mice did not result in any
substantial alteration of their normally expected lifespan, or
showed any visually detectable abnormalities. Although this
does not exclude the need for thorough toxicology experiments,
it may indicate that the dye is not significantly toxic.
5456
www.angewandte.org
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 5452 –5456