Synthesis and evaluation of inhibitors of transthyretin amyloid formation based on the non-steroidal anti-inflammatory drug, flufenamic acid

Article abstract of DOI:10.1016/S0968-0896(99)00066-8

A light scattering-based amyloid fibril formation assay was employed to evaluate potential inhibitors of transthyretin (TTR) amyloid fibril formation in vitro. Twenty nine aromatic small molecules, some with homology to flufenamic acid (a known non-steroidal anti-inflammatory drug) were tested to identify important structural features for inhibitor efficacy. The results of these experiments and earlier data suggest that likely inhibitors will have aromatic-based structures with at least two aromatic rings. The ring or fused ring system occupying the outermost TTR binding pocket needs to be substituted with an acidic functional group (e.g. a carboxylic acid) to interact with complimentary charges in the TTR binding site. The promising TTR amyloid fibril inhibitors ranked in order of efficacy are: 2>4~7>3>9>6>21 (see Structural summary of the best transthyretin amyloid fibril formation inhibitors identified in this study. The order of efficacy is: 2>4~7>3>9>6>21.). Copyright (C) 1999 Elsevier Science Ltd.

Full text of DOI:10.1016/S0968-0896(99)00066-8

Bioorganic & Medicinal Chemistry 7 (1999) 1339±1347
Synthesis and Evaluation of Inhibitors of Transthyretin Amyloid
Formation Based on the Non-steroidal Anti-in¯ammatory Drug,
Flufenamic Acid
Paul W. Baures, Vibha B. Oza, Scott A. Peterson and Je€ery W. Kelly*
Department of Chemistry and The Skaggs Institute of Chemical Biology, The Scripps Research Institute, 10550 North Torrey Pines Road,
MB12, La Jolla, CA 92037, USA
Received 2 November 1998; accepted 2 December 1998
AbstractÐA light scattering-based amyloid ®bril formation assay was employed to evaluate potential inhibitors of transthyretin
(TTR) amyloid ®bril formation in vitro. Twenty nine aromatic small molecules, some with homology to ¯ufenamic acid (a known
non-steroidal anti-in¯ammatory drug) were tested to identify important structural features for inhibitor ecacy. The results of
these experiments and earlier data suggest that likely inhibitors will have aromatic-based structures with at least two aromatic rings.
The ring or fused ring system occupying the outermost TTR binding pocket needs to be substituted with an acidic functional group
(e.g. a carboxylic acid) to interact with complimentary charges in the TTR binding site. The promising TTR amyloid ®bril inhibi-
tors ranked in order of ecacy are: 2>4ꢀ7>3>9>6>21 (see Fig. 5). # 1999 Elsevier Science Ltd. All rights reserved.
Introduction
carrier of T4 in the cerebral spinal ¯uid (CSF). However
in plasma, TTR serves as the backup carrier for T4,
thyroid binding globulin being the primary carrier (KA
6Â109). Our strategy for inhibiting transthyretin amy-
loid ®bril formation is to identify a ligand that will bind
to TTR in human plasma using the largely unoccupied
(ꢀ90% unoccupied) T4 binding site. Ligand binding
stabilizes transthyretin and as a result also increases the
activation barrier associated with the tetramer to
monomeric amyloidogenic intermediate transition, i.e.
the enabling event in amyloid ®bril formation.18 Pre-
vious results from our laboratory demonstrate that
thyroxine is capable of stabilizing the tetrameric form of
transthyretin, preventing amyloid ®bril formation at a
pH below 5.5 where TTR normally self-associates into
amyloid ®brils.15 Thyroid hormone also appears to sta-
bilize TTR in the CSF in vivo, preventing ®bril forma-
tion. Numerous pathological evaluations suggest that
TTR amyloid ®brils are generally not observed in the
brain.15 This is the case even when a particularly
unstable and pathogenic FAP variant such as L55P is
expressed in the CSF, consistent with the role of T4 as a
TTR stabilizing agent in the CSF.
One of the common features of human amyloid disease
is the deposition of insoluble high molecular weight
cross-b-sheet ®brils derived from the self-assembly of
one of 20 human proteins.1±7 In the case of transthyretin
(TTR), the process of amyloid ®bril formation appears
to be the causative agent in senile systemic amyloidosis
(SSA) and familial amyloid polyneuropathy (FAP).7,8
Wild-type TTR can be transformed into amyloid by
partial denaturation (e.g. at reduced pH), a€ording a
monomeric amyloidogenic intermediate with an alter-
natively folded structure that self-assembles into amy-
loid ®brils.1±3,9±15 Transthyretin is a stable tetramer
under physiological conditions; however, a low pH
environment such as that encountered in an endosome
or lysosome (or an analogous non-acidic denaturing
environment) dissociates TTR into monomeric subunits
as well as facilitating a conformational change within
the TTR monomer making it amyloidogenic.16 The
>50 mutations associated with FAP make the kinetics
and thermodynamics of this process more facile.12±14,17
Transthyretin normally binds to the thyroid hormone
thyroxine (1, T4) with negative cooperativity and high
anity (KA1 108; KA2 106), serving as the primary
The stagnant ®bril formation assay developed by our
laboratory11,15 was used to discover the non-steroidal
anti-in¯ammatory drug ¯ufenamic acid (Flu, 2), which
is an excellent TTR ®bril inhibitor.19 Flufenamic acid
binds with high anity and negative cooperativity (pH
7.6) to wild-type (KD1=30‹14 nM, KD2=255‹
97 nM), V30M (KD1=41‹10 nM, KD2=320‹125 nM)
Key words: Transthyretin; amyloid ®bril; inhibition; screening;
NSAID.
* Corresponding author. Tel.: 858-784-9605; fax: 858-784-9610;
e-mail: jkelly@scripps.edu
0968-0896/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S0968-0896(99)00066-8

1340
P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
and L55P TTR (KD1=74‹16nM, KD2=682‹137nM),
completely inhibiting amyloid ®bril formation at a
concentration of 10.8 mM, 3Â the physiological con-
centration of TTR (3.6 mM), under conditions where
TTR amyloid ®bril formation is maximal (pH 4.4).19
A
cocrystal structure of TTR with ¯ufenamic acid was
determined to 2.0 A resolution in collaboration with the
Sacchettini laboratory.19 This structural information
provides the basis for a rational drug design e€ort cen-
tered on identifying the important pharmacophoric
substructures of ¯ufenamic acid, such that subsequent
parallel syntheses could be utilized to make an opti-
mized ligand.
The compounds synthesized and reported in this paper
were designed to test the important structural features
of ¯ufenamic acid that render it a good inhibitor (Fig. 1).
Each of the ®ve molecular fragments of Flu (fragments
A±E) were evaluated in the structure±activity analysis
described herein to di€ering extents. Such information
should eventually lead to a generalizable pharmaco-
phore hypothesis and help identify other novel ligands
as represented by the generic structure on the right side
of Figure 1. In this paper we report ®bril inhibition data
for several NSAIDs including ¯ufenamic acid analogues
utilizing the proven stagnant amyloid ®bril formation
assay.11,15,18 The data suggests that the carboxylic acid
functionality plays an important role in the binding of
amyloid ®bril inhibitors to TTR. In addition, there are
also clear preferences for the aromatic platform that is
capable of making van der Waals interactions with the
TTR binding site.
Figure 2. Structural representation of the non-steroidal antiin-
¯ammatory drugs evaluated as possible inhibitors of transthyretin
amyloid ®bril formation.
meta-tri¯uoromethylaniline (10), and para-amino-
benzoic acid (12) (Fig. 3) were all purchased and used
without further puri®cation. N-Acetyl-meta-tri¯uoro-
methylaniline (11) was prepared by acylation of 10 with
acetic anhydride in pyridine Fig. (3). Compounds 13±15
(Fig. 3) were prepared by reductive amination of the
imine formed by reaction of the appropriate aniline and
benzaldehyde.20 Addition of the appropriate amino-
benzoic acid and tri¯uoromethyl-phenylisocyanate in
THF yielded the urea linked compounds 16±24, whereas
compounds 25±27 were prepared by treating the appro-
priate aminobenzoic acid with phenylisocyanate.21 The
addition of meta-tri¯uoromethylbenzoyl chloride to
either meta- or para-aminobenzoic acid a€ords the ben-
zamides 28 and 29, respectively.
Results
The structures of compounds either purchased or pre-
pared for this study are shown in Figures 2 and 3. The
NSAIDs diclofenamic acid (3), ni¯umic acid (4), indo-
methacin (5), sulindac (6), di¯unisal (7), and tolmetin
(8) (Fig. 2), as well as N-phenylanthranilic acid (9),
The potential inhibitors shown in Figures 2 and 3 were
evaluated in a 72 h stagnant ®bril formation assay
described previously.11,15 Brie¯y, this assay subjects
TTR (3.6 mM) to acidic partial denaturation at pH 4.4
(acetate bu€er, 37 ꢁC) in either the absence (assigned to
be 100% ®bril conversion) or in the presence of a
potential amyloid ®bril inhibitor. The extent of ®bril
inhibition is determined by the optical density measured
on a UV spectrophotometer either at 330 or 400 nM in
the presence of inhibitor relative to the OD reading
in the absence of inhibitor (400 nM monitoring is
employed for drug candidates that absorb signi®cantly
at 330 nM).11,15 A quantitative congo red binding assay
is also performed on promising inhibitors to con®rm the
Figure 1. Flufenamic acid is logically divided into substructures
labeled A±E to identify the portions of the Flu molecule that are
essential for TTR ®bril inhibition. This exercise was performed to
identify the appropriate linker atom or group, R groups and the
appropriate aromatic substructure for an optimal TTR ®bril inhibitor.

P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
1341
Figure 3. Structural representation of the aromatic compounds evaluated as possible inhibitors of transthyretin amyloid ®bril formation. Com-
pounds having yields associated with the compound number were synthesized by our laboratory, the remainder were procured from commercial
sources.
light scattering results, which have proven to be extre-
mely reliable. All potential inhibitors were tested initi-
ally at 36 mM, 10Â the physiological TTR concentration
(3.6 mM), to determine whether there was sucient
activity to warrant further assessment. When comparing
inhibitors it is important to remember that the percent
conversion of soluble transthyretin into ®brils over a
72 h period relative to transthyretin ®bril conversion in
the absence of inhibitor is reported. Hence, 100% ®bril
conversion (®bril formation) re¯ects no inhibition,
whereas 0% conversion is complete inhibition of TTR
®bril formation. We also carried out ¯ufenamic acid
(10.8 mM) inhibition of TTR ®bril formation in each
experiment to calibrate the new inhibitor results (see
next paragraph).
and age of the TTR sample used. In general, the pre-
liminary evaluation of potential inhibitors carried out at
a concentration of 36 mM was done with TTR (3.6 mM)
estimated to be between 90 to 95% pure (SDS-PAGE
analysis). None of these assays were done with TTR less
than 90% pure. In experiments with older TTR (>1
month of age) or protein with ꢀ10% impurities present
or both, we occasionally observed up to a 5% increase
in the extent of apparent amyloid ®bril formation. The
largest contribution to the apparent increase in ®bril
formation (light scattering) in the less pure samples
results from protein impurities that also self-assemble
under amyloid forming conditions, leading to a higher
OD reading at 400 nM. We always employ Flu (2)
(10.8 mM) for the known inhibitor control in our pre-
liminary screening experiments to get a handle on TTR
batch to batch variability. The extent of conversion of
soluble TTR (>90% pure) to amyloid ®brils in the
The exact amount of TTR amyloid ®bril formation
observed from assay to assay is a€ected by the purity

1342
P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
Table 2. Amyloid ®bril inhibition with selected compounds at
varying concentrationsa
presence of 10.8 mM Flu was typically less than 5%. The
range of 1±5% ®bril conversion for Flu inhibition
results from additional light scattering caused by impu-
rities aggregating. We always re-evaluate encouraging
results using >95% purity TTR, which typically yields
only 1% conversion of TTR to ®brils in the presence of
10.8 mM Flu (Table 2).
Concentration (mM)
10.8
3.6
1.8
Compound
Percent ®bril formationb
2
3
4
6
7
9
21
1
2
1
19
1
5
59
26
47
32
67
35
51
86
57
69
62
80
66
72
91
The evaluation of potential inhibitors employing the
®bril formation assay is summarized in Table 1. With
the exception of sulindac (6) and tolmetin (8), the
NSAIDs shown in Figure 2 were generally very good
inhibitors, keeping ®bril conversion to less than 5% of
the control (100%) when evaluated at a concentration
of 36 mM (10Â [TTR]). Sulindac (6) was a good inhi-
bitor, allowing only 14% conversion of TTR to ®brils
relative to the positive control, whereas tolmetin is a
poor inhibitor exhibiting conversion of 75% of TTR
into amyloid ®brils. The remaining compounds shown
in Figure 3 were generally less e€ective, with the excep-
tion of N-phenylanthranilic acid (9) and N-benzyl-para-
aminobenzoic acid (14) which allowed only 2 and 6%
conversion into amyloid ®brils, respectively (36 mM).
Compounds 13, 16, 17 and 21 (Fig. 3) were modest
inhibitors, holding amyloid ®bril conversion to less than
25% of that observed in the positive control at an inhi-
bitor concentration of 36 mM.
aThe TTR concentration was 3.6 mM and the NSAIDs were evaluated
at 10.8 mM, 3.6 mM and 1.8 mM (3, 1 and 0.5 Â the TTR concentration,
respectively).
b Control ®bril formation was determined in each assay in the absence
of drug and was considered to be 100%.
(1Â[TTR]) and 1.8 mM (0.5Â[TTR]), facilitating a direct
comparison to the established inhibitor ¯ufenamic acid,
2.18,19 In addition to light scattering, the extent of amy-
loid ®bril formation was also evaluated using a quanti-
tative congo red binding analysis, congo red being a dye
that exhibits selective binding to amyloid ®brils.22 The
results of this assay are summarized in Table 2 and
illustrated in bar graph format in Figure 4. Flufenamic
acid (2) at a concentration of 10.8 mM, or 3Â the TTR
concentration, is capable of keeping ®bril formation to
less than 1% of that observed in the positive control.
The ®bril conversion yield increases to 26% and 57% as
the concentration of ¯ufenamic acid is reduced to
3.6 mM and 1.8 mM, respectively. Diclofenamic acid (3),
ni¯umic acid (4), and N-phenylanthranilic acid (9) have
the greatest homology to 2. NSAID 4 is nearly as good
as Flu (2) as an amyloid ®bril inhibitor. The NSAIDs 3
and 9 exhibit a bit less than a two fold drop in activity
relative to Flu at an inhibitor concentration of 3.6 mM.
Sulindac (6) and 21 were not nearly as e€ective at these
concentrations.
Compounds 3, 4, 6, 7, 9, and 21 were evaluated in
triplicate using highly puri®ed wild-type TTR at inhi-
bitor concentrations of 10.8 mM (3Â[TTR]), 3.6 mM
Table 1. Amyloid ®bril inhibition by the compounds shown in Figure
1a
Compound
Percent ®bril formationb
3
4
3
2
5
3
6
7
14
3
8
9
74
2
Of the NSAIDs that are not of the anthranilic acid
class, di¯unisal (7) gave the best results. Di¯unisal is a
very promising compound, and the biphenyl sub-
structure an appealing platform to develop a new class
of TTR amyloid ®bril inhibitors. At a concentration of
10.8 mM, inhibitor 7 is indistinguishable from inhibitors
2 and 4, all three exhibiting less than 1% conversion to
amyloid ®brils. At an equal inhibitor to protein con-
centration ratio (3.6 mM) 7 is very similar to 2 and 4,
a€ording 35%, 26% and 32% ®bril formation conver-
sion yields, respectively. The promising inhibitors tested
at a concentration of 3.6 mM ranked in order of ecacy
are: 2>4ꢀ7>3>9>6>21 (Fig. 5). This order of e€ec-
tiveness is also maintained at an inhibitor concentration
of 1.8 mM. The congo red results, which report on the
extent of TTR ®bril formation as a function of inhibitor
concentration parallel the results derived from light
scattering at 400 nM (Fig. 4). Negative values are some-
times observed in the congo red binding analysis when
®bril formation is minimal, such as the case for inhibi-
tors 2, 3, 4, 7 and 9 when evaluated at a concentration
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
100
100
100
15
6
36
22
18
39
35
52
20
100
75
50
55
81
57
83
86
a The TTR concentration was 3.6 mM and the compounds were eval-
uated at 36 mM.
b Control ®bril formation was determined in each assay in the absence
of drug and was assigned to be 100%.

P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
1343
Figure 4. A bar graph representing the extent of TTR amyloid ®bril formation as a function of inhibitor concentration at constant TTR con-
centration (3.6 mM). Dark bars represent ®bril formation as evaluated by the optical density (OD) reading of the solution in a UV spectrometer at
400 nM. Lighter bars represent the extent of TTR ®bril formation as evaluated by congo red binding. Each assay was done in triplicate and the
error bars indicate the standard deviation. Asterisks replacing the congo red determinations indicate small negative values for congo red binding.
Occasionally small negative values are observed when little or no ®bril formation occurs.
19
.
of 36 mM, re¯ecting the greater error involved in this
measurement and possibly a systematic error (Fig. 4).
of the Flu2 TTR complex. Owing to the symmetry
associated with the T4 binding site, Flu is able to bind in
two symmetry equivalent modes in each of two sym-
metry equivalent binding sites shown in Figure 6. In
addition, Flu is able to bind in two conformations
where the CF3 group is either in a cis or trans relation-
ship to the intramolecularly hydrogen bonded NH and
COOH functional groups (Fig. 6). The crystal structure
Discussion
A collaboration between the Sacchettini laboratory
and our own produced a 2 A resolution crystal structure
Figure 5. Structural summary of the best transthyretin amyloid ®bril formation inhibitors identi®ed in this study. The order of ecacy is:
2>4ꢀ7>3>9>6>21.

1344
P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
hysteresis characteristic of TTR GdnHCl mediated
unfolding and reconstitution curves.12 The other
challenge associated with inhibiting the amyloidogeni-
city of transthyretin is its high plasma concentration
(3.6 mM), necessitating high inhibitor concentrations
3.6±7.2 mM.18
As illustrated in Figure 1, we have segregated the struc-
ture of 2 into ®ve groups to focus on these substructures
and evaluate their importance for inhibitor ecacy.
Potential inhibitors were purchased and synthesized
(Figs. 2 and 3) to decipher which of the structural fea-
tures of Flu (the tri¯uoromethyl group, the aniline
nitrogen linker, the carboxylic acid group, or the diaryl
system) are most important. For example, diclofenamic
acid, 3, retains the diarylamine skeleton of 2, but has an
alternative substitution pattern on the aromatic rings
and possibly an altered conformation(s). Ni¯umic acid,
4, is identical to Flu except for the heterocyclic aromatic
ring. The remaining NSAIDs 5±8 bear less resemblance
to Flu, whereas N-phenylanthranilic acid, 9, is basically
Flu without the tri¯uoromethyl group. The impressive
ecacy of compound 9 as an amyloid ®bril inhibitor
questions the importance of the tri¯uoromethyl group
identi®ed in crystallographic studies as a key pharma-
cophore in the activity of Flu. However, a comparison
of inhibitors 13 to 14 (at 10.8 mM, data not shown) and
distinguishing between the inhibitor ecacy of tri¯uoro-
substituted aryl ureas to the unsubstituted aryl ureas
suggests that the CF3 group is bene®cial (Table 1 and
Fig. 3). Future structural, binding and kinetic studies
comparing 9 to Flu and analogous comparisons should
prove particularly important for further deliniating the
importance of the tri¯uoromethyl group.
Figure 6. A schematic representation of the cocrystal structure
between transthyretin and ¯ufenamic acid. There are two symmetry
equivalent ¯ufenamic acid binding sites/tetramer, each of which can
accommodate ¯ufenamic acid in two symmetry equivalent modes.
Flufenamic acid also exists in two conformations (i.e. the CF3 sub-
stituent either being cis or trans to the intramolecular hydrogen bonded
NH and COO groups). A more detailed description of this structure
can be found in the literature.19
also clearly shows the interaction between the Lys-15
ammonium group from each TTR subunit and the
COO group of Flu (Fig. 6). Interestingly, the CF3
group of Flu is placed in a region of TTR that is quite
hydrophilic in the apo structure, facilitating a structural
rearrangement in TTR to accomodate the CF3 sub-
stituent of Flu. In the apo structure, the side chains of
all four Ser-117 residues are pointing towards the two
T4 binding cavities interacting with bulk H2O ®lling
the empty binding site. Accomodation of the CF3 from
Flu in the T4 site causes the side chains of the Ser-117
residues on all four subunits to rotate approximately
120ꢁ, forming two non-solvated hydrogen bonds
between the Ser-117 residues on adjacent subunits (four
new hydrogen bonds/tetramer) (Fig. 6). These inter-
subunit hydrogen bonds could be important in further
stabilizing the tetramer and possibly in making the o€
rate for inhibitor dissociation slow.
Discovering the e€ectiveness of di¯unisal (7) is arguably
the most important result to come from this study.
NSAID 7 is nearly equivalent to inhibitor 4 and is
nearly as good as ¯ufenamic acid (2) in the stagnant
®bril formation assay employed in these studies. Before
we can truly understand the SAR rami®cations of these
results with regard to Flu, a structure of the di¯uni-
.
.
sal TTR complex and perhaps other inhibitor TTR
complexes are required, as it is likely that these aromatic
compounds bind di€erently to TTR.
A number of compounds were synthesized varying the
linker separating the two aryl rings. These linkers
included the aminomethylene group in 13±15, the urea
linkers in compounds 16±27, and the amides 28 and 29.
Incorporation of each of these linkers results in di€erent
conformational biases, separation geometries and dis-
tances relative to the substituted aryl rings. By alternat-
ing the positions of the tri¯uoromethyl and carboxylic
acid functionalities, we were able to search a variety of
conformational space in hopes of identifying combina-
tion(s) that might be best for optimizing binding of
these `linker modi®ed' compounds to TTR. Structures
10±12 retain only a single aryl ring and were included in
the bioassay portion of the experiments as controls and
were inactive. The potential inhibitors 13, 14 and 15
with the aminomethylene linker exhibit biological
activity at an inhibitor concentration of 36 mM, but do
Inhibiting TTR conformational changes with small
molecules is di€erent than enzyme inhibition. Enzymes
have the opportunity to rebind their inhibitors after
they dissociate, whereas apo-TTR under amyloid form-
ing conditions changes conformation and associates,
denying the alternatively folded protein another chance
to bind the inhibitor. The activation barrier(s) for TTR
refolding are high enough to prevent the associated
alternatively folded protein to be in equilibrium with
folded TTR, even in the presence of high concentrations
of high anity inhibitors. That the reconstitution pro-
cess has associated with it a high ÁG{ is re¯ected in the

P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
1345
not exhibit signi®cant biological activity at 10.8 mM
concentration (data not shown). Of the urea linked
compounds, only 16, 17 and 21 showed inhibitor e-
cacy at a concentration of 36 mM. The remainder of the
compounds mentioned in this paragraph do not exhibit
impressive inhibitory properties at a concentration of
36 mM. It is tempting to speculate that the spacer (Fig. 1)
is not so important, based on the observation that
compounds with a variety of spacers, including a direct
aryl linkage (e.g. 7), serve as inhibitors. The ability of the
TTR binding site to accomodate a number of di€erent
types of spacers is also in keeping with the results from
another recent study which demonstrates that appro-
priately substituted stilbenes, dibenzofurans and biphe-
nylethers are also good inhibitors.18 Although caution is
appropriate regarding drawing conclusions about the
spacers as it is likely that the bound structures of these
compounds with TTR will di€er.
screens suggest that likely inhibitors will have aromatic-
based structures with at least two aromatic rings (one of
the two rings can be a bi- or tricyclic aromatic ring18).
The ring or fused ring system18 occupying the outermost
binding pocket of TTR needs to be substituted with an
acidic functional group, either a phenol or carboxylic
acid (Fig. 6). The promising inhibitors identi®ed in this
study ranked in order of ecacy are: 2>4ꢀ7>3
>9>6>21 (Fig. 5).
Experimental
General aspects
All glassware were oven-dried and cooled in a dessicator
containing CaSO4. THF was distilled from Na/benzo-
phenone. Anhydrous methanol was obtained by distilling
HPLC grade methanol from magnesium methoxide which
was stored over activated 4 A molecular sieves. Thin-
layer chromatography was performed on Kodak plastic
backed silica gel plates 250 m that were visualized by UV
irradiation or I2 staining or both. Chromatographic
puri®cation on silica gel (Merck, grade 60, 240±400
mesh, 60 A) was done by ¯ash chromatography. 1H
NMR spectra were recorded at 300 or 400 MHz and 13C
NMR at 75 MHz using either Bruker or Varian spec-
trometers. A pulse delay of at least 2 s was employed to
avoid losing signals for quaternary carbons. Most of the
aromatic inhibitors characterized by 13C NMR exhibit
7±13 resonances, slightly less that the maximal number
expected ꢀ15, most likely due to resonance overlap.
DMSO-d6 (1H d 2.49, m, 13C d 39.5, m) or CD3OD (1H
d 4.87, s, 3.31, m; 13C d 49.15, m) were used as interal
standards, coupling constants (J) values are reported in
Hertz.
What is striking regarding the 29 compounds evaluated
within is that the aromatic carboxylic acid or phenol
functional group is important, but not sucient (e.g. 12,
Fig. 3 and Table 1) for TTR ®bril inhibitor ecacy. In
.
the cocrystal structure of thyroxine TTR, the carboxyl
and amino groups of T4 form ion pairs with Lys-15 and
Glu-54 on the solvent exposed region of the binding
cavity.23,24 An ion pair is also observed between the
carboxylate of ¯ufenamic acid and the two Lys-15 e-
ammonium groups in its cocrystal structure with TTR
(Fig. 6).19 The results from these studies infer that the
formation of a tight ion pair at the solvent exposed
region of the binding cavity plays an important role in
determining the ability of the compounds to bind to
TTR. This trend was also observed in another analysis
of 85 commercially available compounds.18
The amyloid ®bril inhibitory activity of 2, 3, 4, 7, 9 and
14, as well as the modest activity observed in several of
the urea analogues (most noteably 16, 17 and 21) indi-
cates that signi®cant di€erences in the structural plat-
form (aromatic template) are tolerated in the inhibitors.
It is too early to be certain what the best structural
platform is for a TTR amyloid ®bril inhibitor. We are
currently using these results as an incentive for testing a
number of other commercially available and syntheti-
cally accessible compounds to identify new inhibitors or
aromatic templates or both. The compounds selected
for future screens will be aromatics with variable sub-
stituents containing a variety of linkers. Likely inhibi-
tors will also contain a carboxylic acid, phenolic, or
equivalent group expected to interact with one or more
Lys-15 in TTR. A zwitterionic substituent will also be
evaluated to interact with TTR analogously to T4.
General procedure for reductive amination reactions to
yield 13±15
To a round bottom ¯ask was added the appropriate
aniline (6.21 mmol), the benzaldehyde (6.21 mmol),
and 2.7 mol % of acetic acid in approximately 30 mL
of benzene. The solution was re¯uxed in a Dean±Stark
apparatus for 2±24 h, the duration being dependent on
the nature of the aldehyde used in the reaction. The
solvent was removed by reducing the volume under
re¯ux and then by rotary evaporation under reduced
pressure. The Schi€ base was dissolved in 5 mL of
anhydrous methanol and cooled to 0 ꢁC. To this stirred
solution was added 0.39 g of sodium cyanoborohydride
(6.21 mmol) in small portions over 5 min (vigorous
bubbling was observed after the addition of solid
sodium cyanoborohydride). The reaction vessel was
vented to the atmosphere with a 21 gauge needle
through a septa. The reaction mixture was stirred for 18 h
and the methanol removed by rotary evaporation under
reduced pressure. The residual material was parti-
tioned between 50 mL of EtOAc and 5 mL of distilled
H2O and the organic layer was washed with 5 mL of
brine solution before drying it over anhydrous MgSO4.
The solution was concentrated under vacuum and pur-
i®ed by either crystallization or silica gel chromatography.
Conclusion
We have utilized the stagnant ®bril formation assay to
evaluate potential inhibitors of TTR amyloid ®bril for-
mation in vitro. Twenty nine aromatic small molecules,
some with homology to ¯ufenamic acid, were tested to
identify important structural features for inhibitor e-
cacy. The results of these experiments and earlier

1346
P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
Summary of spectral data
7.56 Hz, 1H); 13C NMR (75 MHz, DMSO-d6): d 167.40,
152.55, 144.06, 136.27, 133.02, 130.93, 126.12, 124.30,
124.09, 117.57; HRMS (FAB) calcd for C15H11F3N2O3
325.0800, found 325.0804 (M+H)+.
13: 1H NMR (300 MHz, DMSO-d6): d 7.96 (d, J=
7.9 Hz, 2H), 7.5 (d, J=7.9 Hz, 2H), 7.25 (dd, J=8,
7.8 Hz), 6.85 (m, 4H, including ArH and NH), 4.43 (d,
J=3.9 Hz, 2H); 13C NMR (75 MHz, CD3OD): d 150.38,
146.72, 131.17, 130.80, 128.26, 116.79, 113.884, 109.96,
48.04 HRMS (FAB) calcd for C15H12F3NO2 296.0898,
found 296.0897 (M+H)+.
1
19: H NMR (300 MHz, DMSO-d6): d 10.53 (s, 1H),
10.18 (s, 1H), 8.41 (d, J=8.44 Hz, 1H), 8.04 (s, 1H),
7.23 (d, J=8.08 Hz, 1H), 7.72 (d, J=8.84 Hz, 1H), 7.50
(dd, dd, J=7.36, 9.2, 8.44, 8.08 Hz, 2H), 7.25 (d,
J=7.72 Hz, 1H), 7.03 (dd, J=7.72, 7.36 Hz, 1H); 13C
NMR (75 MHz, DMSO-d6): d 169.71, 152.38, 142.13,
140.85, 133.92, 131.18, 129.89, 122.13, 121.27, 119.96,
188.34, 116.46, 115.67, 114.67; HRMS (FAB) calcd for
C15H11F3N2O3 325.0800, found 325.0804 (M+H)+.
14: 1H NMR (300 MHz, CD3OD): d 7.73 (d, J=
8.84 Hz, 2H), 7.45±7.35 (m, 5H), 6.57 (d, J=8.84 Hz,
2H), 4.36 (s, 2H); 13C NMR (75 MHz, CD3OD): d
155.5, 132.74, 129.64, 128.32, 128.11, 1112.55, 47.89;
HRMS (FAB) calcd for C14H13NO2 228.1025, found
228.1024 (M+H)+.
20: 1H NMR (300 MHz, DMSO-d6): d 9.06 (s, 1H), 9.01
(s, 1H), 8.20 (s, 1H), 8.04 (s, 1H), 7.66 (d, J=8.0 Hz,
1H), 7.59 (d, J=5.12 Hz, 1H), 7.58 (d, J=4.6 Hz, 1H),
7.45 (dd, J=7.56, 8.08 Hz, 1H), 7.38 (dd, J=7.84,
8.08 Hz, 1H), 7.25 (d, J=7.0 Hz, 1H); 13C NMR
(75 MHz, DMSO-d6): d 167.77, 152.95, 140.88, 140.10,
131.82, 130.14, 129.38, 123.44, 123.06, 122.28, 119.62,
118.51, 114.72; HRMS (FAB) calcd for C15H11F3N2O3
325.0800, found 325.0805 (M+H)+.
15: 1H NMR (300 MHz, DMSO-d6): d 7.94 (d,
J=7.8 Hz, 2H), 7.48 (d, J=8.1 Hz, 2H), 7.35 (t,
J=5.6 Hz, 1H), 7.13 (s, 2H), 7.15 (s, 1H); 13C NMR
(75 MHz, CD3OD): d 151.14, 145.51, 133.66, 133.23,
131.26, 128.26, 112.84, 109.74, 47.78; HRMS (FAB)
calcd for C16H11F6NO2 364.0772, found 364.0779
(M+H)+.
21: 1H NMR (300 MHz, DMSO-d6): d 9.14 (s, 1H), 9.12
(s, 1H), 8.01 (s, 1H), 7.89 (d, J=8.64 Hz, 2H), 7.59 (d,
J=8.64 Hz, 2H), 7.56 (s, 1H), 7.46 (dd, J=7.48 Hz,
1H), 7.26 (d, J=7.56 Hz, 1H); 13C NMR (75 MHz,
DMSO-d6): d 167.59, 152.61, 144.06, 140.59, 130.95,
130.05, 129.90, 124.41, 122.26, 118.65, 117.87, 114.78;
HRMS (FAB) calcd for C15H11F3N2O3 325.0800, found
325.0804 (M+H)+.
General procedure for urea formation reactions to a€ord
16±27
A round-bottom ¯ask was charged with the appropriate
aniline in 5 mL of anhydrous THF under N2. The solu-
tion was cooled to 0 ꢁC and the appropriate isocyanate
was added dropwise over 30 s. The reaction was allowed
to warm to room temperature. In general, a heavy pre-
cipitate forms within 5±60 min. In the case of the more
sterically constrained systems bearing ortho substituents,
the reaction may require up to 24 h to precipitate. The
products were collected by vacuum ®ltration and the
solids washed with 25±50 mL of dichloromethane before
drying the urea under vacuum.
1
22: H NMR (300 MHz, DMSO-d6): d 10.54 (s, 1H),
10.22 (s, 1H), 8.41 (d, J=7.72 Hz, 1H),), 7.97 (d, J=
7.9 Hz, 1H), 7.75 (d, J=8.46 Hz, 2H), 7.59 (d, J=
8.82 Hz, 2H), 7.51 (dd, J=7.72, 7.54 Hz, 1H), 7.02 (dd,
J=7.88, 7.74 Hz, 1H); 13C NMR (75 MHz, DMSO-d6):
d 169.74, 152.32, 143.79, 142.15, 133.89, 131.22, 126.08,
126.05, 121.31, 120.09, 118.44, 115.84; HRMS (FAB)
calcd for C15H11F3N2O3 325.0800, found 325.0806
(M+H)+.
Summary of spectral data
1
16: H NMR (300 MHz, DMSO-d6): d 10.31 (s, 1H),
9.19 (s, 1H), 8.26 (d, J=8.64 Hz, 1H), 7.85 (d, J=
8.12 Hz, 1H), 7.61 (d, J=7.84 Hz, 1H), 7.54 (dd, 2H),
7.41 (dd, J=7.6 Hz, 1H), 7.3 (dd, J=6.5 Hz, 1H), 6.93
(dd, J=7.48 Hz, 1H); 13C NMR (75 MHz, DMSO-d6): d
169.45, 153.66, 142.15, 136.07, 133.78, 133.11, 131.12,
130.11, 126.46, 126.41, 126.13, 121.25. 120.26, 116.21;
HRMS (FAB) calcd for C15H11F3N2O3 325.0800, found
325.0804 (M+H)+.
1
23: H NMR (300 MHz, DMSO-d6): d 9.09 (s, 1H), 9.0
(s, 1H), 8.20 (s, 1H), 7.70±7.55 (6d/m, 6H), 7.38 (dd,
J=7.84 Hz, 1H); 13C NMR (75 MHz, DMSO-d6): d
167.60, 152.55, 143.55, 139.87, 131.67, 129.16, 126.21,
126.16, 123.32, 122.84, 119.44, 118.18; HRMS (FAB)
calcd for C15H11F3N2O3 325.0800, found 325.0804
(M+H)+.
17: 1H NMR (300 MHz, DMSO-d6): d 9.61 (s, 1H), 8.17
(dd, J=1.88 Hz, 1H), 8.08 (s, 1H), 7.96 (d, J=8.08 Hz,
1H), 7.65 (m, 4H), 7.40 (dd, J=8.12, 7.84 Hz, 1H), 7.24
(dd, J=7.56 Hz, 1H); 13C NMR (75 MHz, DMSO-d6): d
167.39, 152.58, 139.89, 136.30, 132.98, 131.56, 129.24,
126.04, 125.92, 125.81, 123.88, 123.09, 122.36, 118.96;
HRMS (FAB) calcd for C15H11F3N2O3 325.0800, found
325.0803 (M+H)+.
24: 1H NMR (300 MHz, DMSO-d6): d 9.17 (s, 1H), 9.14
(s, 1H), 7.93 (dd, J=8.64 Hz, 2H),), 7.67 (dd, J=
8.36 Hz, 2H), 7.59 (m, 4H); 13C NMR (75 MHz,
DMSO-d6): d 167.42, 152.28, 143.90, 143.37, 130.85,
126.26, 126.22, 124.33, 118.25, 117.69; HRMS (FAB)
calcd for C15H11F3N2O3 325.0800, found 325.0803
(M+H)+.
1
25: H NMR (300 MHz, DMSO-d6): d 10.45 (s, 1H),
9.83 (s, 1H), 8.43 (d, J=8.64 Hz, 1H),), 7.99 (d, J=
7.8 Hz, 1H), 7.59 (d, J=7.84 Hz, 2H), 7.57 (dd,
18: 1H NMR (300 MHz, DMSO-d6): d 9.72 (s, 1H), 8.19
(s, 1H), 7.93 (m, 3H), 7.6 (m, 4H), 7.23 (dd, J=7.32,

P. W. Baures et al. / Bioorg. Med. Chem. 7 (1999) 1339±1347
1347
J=7.8 Hz, 2H), 6.95 (dd, dd, J=7.84 Hz, 2H); 13C
NMR (75 MHz, DMSO-d6): d 169.73, 152.55, 142.54,
139.96, 133.89, 131.21, 128.87, 122.29, 120.99, 120.08,
119.03; HRMS (FAB) calcd for C14H12N2O3 257.0848,
found 257.0851 (M+H)+.
The compounds were originally prepared as 5.4 mM
solutions in pure DMSO such that only 2 mL of solution
was required to reach the desired concentration (36 mM,
10Â TTR concentration) in a 300 mL assay. In the case
of the lower concentrations of inhibitors (10.8 mM,
3.6 mM TTR, 1.8 mM), serial dilutions were made of the
36 mM inhibitor solutions such that 2 mL provided the
desired amount of compound to be tested. All of the
compounds were prepared by weighing a known
amount of material and making the appropriate addi-
tion of DMSO. The remainder of the assay was per-
formed as described, except for inhibitors that absorbed
at 330 nM, in which case 400 nM light was employed for
the optical density measurements.
1
26: H NMR (300 MHz, DMSO-d6): d 8.9 (s, 1H), 8.69
(s, 1H), 8.19 (dd, J=1.88 Hz, 1H), 7.69 (d, J=7.82 Hz,
1H), 7.59 (d, J=7.84 Hz, 1H), 7.45 (d, J=7.8 Hz, 2H),
7.39 (dd, J=7.82 Hz, 1H), 7.25 (dd, J=7.85 Hz, 2H),
6.98 (dd, J=6.2 Hz, 1H); 13C NMR (75 MHz, DMSO-
d6): d 167.58, 152.73, 140.19, 139.74, 131.56, 129.17,
128.97, 122.93, 122.58, 122.18, 119.12, 118.57; HRMS
(FAB) calcd for C14H12N2O3 257.0848, found 257.0850
(M+H)+.
27: 1H NMR (300 MHz, DMSO-d6): d 9.04 (s, 1H), 8.76
(s, 1H), 7.98 (d, J=8.36 Hz, 2H), 7.65 (d, J=8.36 Hz,
2H), 7.51 (d, J=7.8 Hz, 2H), 7.26 (dd, J=7.56, 7.28 Hz,
2H), 6.94 (dd, J=7.04 Hz, 1H); 13C NMR (75 MHz,
DMSO-d6): d 167.72, 152.72, 144.49, 139.76, 131.85,
129.22, 124.16, 122.62, 118.93, 117.71; HRMS (FAB)
calcd for C14H12N2O3 257.0921, found 257.0926
(M+H)+.
Acknowledgements
We thank the National Institutes of Health (NIH R01
DK46335), the Lita Annenberg Hazen Family, the
Skaggs Family, and the NIH-NRSA to P.B. (F32
GM18451-01) for partial funding of this work.
References
1. Kelly, J. W. Curr. Op. Struct. Biol. 1996, 6, 11.
2. Kelly, J. W. Structure 1997, 5, 595.
General procedure for benzamide forming reactions to
a€ord 28 and 29
3. Kelly, J. W. Curr. Op. Struct. Biol. 1998, 8, 101.
4. Lansbury, P. T. Biochemistry 1992, 31, 6865.
5. Sipe, J. D. Ann. Rev. Biochem. 1992, 61, 947.
6. Sipe, J. D. Crit. Rev. in Clin. Lab. Sci. 1994, 31, 325.
7. Jacobson, D. R.; Buxbaum, J. N. Adv. Hum. Genet. 1991,
20, 69.
A round-bottom ¯ask was charged with the appropriate
aniline (7.3 mmol) in 5 mL of anhydrous THF under N2.
The solution was cooled to 0 ꢁC and meta-tri-
¯uoromethylbenzoyl chloride (3.3 mmol) was added
dropwise over 30 s. The reaction was allowed to warm
to room temperature. After 24 h, the solution was par-
titioned between 50 mL EtOAc and 40 mL of 1 N HCl.
The organic layer was washed with H2O and brine,
dried over anhydrous MgSO4, and concentrated under
reduced pressure. The resulting solids were crystallized
from MeOH:EtOAc:hexanes.
8. Benson, M. D. TIBS 1989, 12, 88.
9. Colon, W.; Kelly, J. W. Biochemistry 1992, 31, 8654.
10. Kelly, J. W.; Lansbury, P. T. J. Amyloid: Int. J. Exp. Clin.
Invest. 1994, 1, 186.
11. Lai, Z.; Colon, W.; Kelly, J. W. Biochemistry 1996, 35,
6470.
12. Lai, Z.; McCulloch, J.; Kelly, J. W. Biochemistry 1997, 36,
10230.
13. McCutchen, S.; Colon, W.; Kelly, J. W. Biochemistry
1993, 32, 12119.
14. McCutchen, S. L.; Lai, Z.; Miroy, G.; Kelly, J. W.; Colon,
W. Biochemistry 1995, 34, 13527.
15. Miroy, G. J.; Lai, Z.; Lashuel, H.; Peterson, S. A.; Strang,
C.; Kelly, J. W. Proc. Natl. Acad. Sci. USA 1996, 93,
15051.
16. Lashuel, H. A.; Lai, Z.; Kelly, J. W. Biochemistry 1998,
37, 17851.
Summary of spectral data
28: 1H NMR (300 MHz, DMSO-d6): d 8.43 (dd, J=
1.88 Hz, 1H), 8.33 (s, 1H), 8.28 (d, J=7.84 Hz, 1H), 8.08
(d, J=7 Hz, 1H), 7.91 (d, J=7.84 Hz, 1H), 7.72 (dd,
J=7.84 Hz, 2H), 7.47 (dd, J=8.08, 7.84 Hz, 1H); 13C
NMR (75 MHz, CD3OD): d 167.35, 164.28, 139.24,
135.59, 132.02, 131.41, 129.79, 129.05, 128.34, 124.92,
124.66, 124.47, 124.43 121.44; HRMS (FAB) calcd for
C15H10F3NO3 310.0691, found 310.0687 (M+H)+.
17. Inlow, J.; Lai, V.; Kelly, J. W. Biochemistry 1998, 37, in press.
18. Baures, P. W.; Peterson, S. A.; Kelly, J. W. Bioorg. Med.
Chem. Lett. 1998, 6, 1389.
19. Peterson, S. A.; Klabunde, T.; Lashuel, H. A.; Purkey, H.;
Sacchettini, J. C.; Kelly, J. W. Proc. Natl. Acad. Sci. USA
1998, 95, 12956.
20. Borch, R. F.; Bernstein, M. D.; Durst, H. D. J. Am.
Chem. Soc. 1971, 93, 2897.
21. Satchell, D. P. N.; Satchell, R. S. Chem. Soc. Rev. 1975, 4,
231.
22. Klunk, W. E.; Pettegrew, J. W.; Abraham, D. J. J. His-
tochem. Cytochem. 1989, 37, 1273.
29: 1H NMR (300 MHz, DMSO-d6): d 8.29 (s, 1H), 8.26
(d, J=8.08 Hz, 1H), 7.93 (m, 5H), 7.78 (dd, J=7.84 Hz,
1H); 13C NMR (75 MHz, CD3OD): d 166.96, 164.54,
142.98, 132.06, 130.34, 129.85, 128.44, 125.86, 124.47,
119.74, 119.65; HRMS (FAB) calcd for C15H10F3NO3
310.0691, found 310.0688 (M+H)+.
23. Wojtczak, A.; Luft, J.; Cody, V. J. Biol. Chem. 1992, 267,
353.
Three day stagnant ®bril forming assay
The compounds were tested by following the previously
published protocol incorporating the following changes.4
24. Wojtczak, A.; Cody, V.; Luft, J. R.; Pangborn, W. Acta
Cryst. 1996, D52, 758.