Structure-based design of N-phenyl phenoxazine transthyretin amyloid fibril inhibitors

Article abstract of DOI:10.1021/ja993309v

Starting with the published 2.0 A X-ray crystal structure of the transthyretin.(flufenamic acid)₂ complex, a simple structure-based ligand design strategy was employed to conceive of N-phenyl phenoxazine transthyretin (TTR) amyloid fibril inhib

Full text of DOI:10.1021/ja993309v

2178
J. Am. Chem. Soc. 2000, 122, 2178-2192
Structure-Based Design of N-Phenyl Phenoxazine Transthyretin
Amyloid Fibril Inhibitors
H. Michael Petrassi,‡ Thomas Klabunde,† James Sacchettini,† and Jeffery W. Kelly*,‡
Contribution from the Department of Chemistry and The Skaggs Institute of Chemical Biology,
The Scripps Research Institute, 10550 North Torrey Pines Rd, MB12, La Jolla, California 92037, and
Department of Biochemistry and Biophysics, Texas A&M UniVersity, College Station, Texas 77843
ReceiVed September 13, 1999
Abstract: Starting with the published 2.0 Å X-ray crystal structure of the transthyretin‚(flufenamic acid)2
complex, a simple structure-based ligand design strategy was employed to conceive of N-phenyl phenoxazine
transthyretin (TTR) amyloid fibril inhibitors. Fifteen N-phenyl phenoxazines were chemically synthesized and
evaluated using a quantitative amyloid fibril assay in vitro. The structure of one of the two most active
phenoxazines, 4, bound to TTR was solved to a resolution of 1.9 Å to understand the structural basis of its
efficacy. N-phenyl phenoxazine 4 binds similar to the orientation anticipated, although not as deeply into the
channel as expected. Like flufenamic acid, 4 mediates binding-induced conformational changes that enable
intersubunit H-bonding in tetrameric TTR which may be important for preventing fibril formation. Analytical
ultracentrifugation analysis demonstrates that 4 blocks the first step of TTR amyloid fibril formation, that is,
tetramer dissociation to the alternatively folded amyloidogenic monomer. Isothermal titration calorimetry was
used to determine the binding constants of 4 to TTR and to dissect the enthalpy and entropy contributions
associated with ligand binding. Phenoxazine 4 exhibits binding and inhibitor efficacy against WT TTR that is
very similar to that of flufenamic acid, unlike the situation with the inhibition of L55P fibril formation where
4 is superior to Flu as an inhibitor but not as a binder. It is clear that 4 functions in part by stabilizing the
normally folded tetramer through formation of the TTR‚(4)2 complex, which in turn increases the activation
energy for tetramer dissociation. The data also suggest that 4 destabilizes the transition state associated with
TTR dissociation to the monomeric amyloidogenic intermediate. Future biophysical studies, including kinetic
measurements, are needed to understand the exact mechanism(s) of the action of 4.
Transthyretin (TTR) functions as the backup carrier of
of age) associated with cardiac dysfunction,12,13 whereas, fibril
formation from one of >70 single site variants of TTR is
genetically linked to pathology in FAP, which refers to a
collection of early onset diseases characterized by neurodegen-
eration and/or organ dysfunction.14-21
Transthyretin is a 127-residue, â-sheet rich, homotetrameric
protein (3.6 µM in plasma) that dissociates under acidic
conditions from its normal quaternary structure to an alterna-
tively folded monomeric amyloidogenic intermediate. 7-11,22-25
This intermediate self-assembles into protofilaments, filaments,
and, under certain conditions, amyloid fibrils that render the
thyroxine (T4) in plasma and the main carrier of retinol binding
protein.1-6 In unfortunate individuals, plasma TTR is converted
into amyloid fibrils in the extracellular space via misfolding
and self-assembly, a process we have been able to simulate in
vitro.7-11 The self-assembly of transthyretin into amyloid fibrils
is implicated as the cause of two amyloid diseases, senile
systemic amyloidosis (SSA) and familial amyloid polyneur-
opathy (FAP). Wild-type (WT) transthyretin deposition is the
putative cause of SSA, a late onset amyloid disease (80 years
* To whom correspondence should be addressed. Telephone: (858) 784-
9605. Fax: (858) 784-9610. E-mail: jkelly@scripps.edu.
† Texas A&M University.
(12) Cornwell, G. C.; Sletten, K.; Johansson, B.; Westermark, P. Biochem.
Biophys. Res. Comm. 1988, 154, 648-653.
‡ The Scripps Research Institute.
(1) Blake, C. C. F.; Geisow, M. J.; Oatley, S. J. J. Mol. Biol. 1978, 121,
339-356.
(2) Monaco, H. L.; Rizzi, M.; Coda, A. Science 1995, 268, 1039-1041.
(3) Pages, R. A.; Robbins, J.; Edelhoch, H. Biochemistry 1973, 12, 2773-
2780.
(13) Westermark, P.; Sletten, K.; Johansson, B.; Cornwell, G. G. Proc.
Natl. Acad. Sci. U.S.A. 1990, 87, 2843-2845.
(14) Jacobson, D. R.; Buxbaum, J. N. AdV. Hum. Genet. 1991, 20, 69-
123.
(15) Jacobson, D. R.; Pastore, R. D.; Yaghoubian, R.; Kane, I.; Gallo,
G.; Buck, F. S.; Buxbaum, J. N. N. Engl. J. Med. 1997, 336, 466-73.
(16) Benson, M. D. TIBS 1989, 12, 88-92.
(4) Ferguson, R. N.; Edelhoch, H.; Saroff, H. A.; Robbins, J. Biochem.
1975, 14, 282-289.
(5) Wojtczak, A.; Luft, J.; Cody, V. J. Biol. Chem. 1992, 267, 353-
(17) Coelho, T. Curr. Opin. Neurol. 1996, 9, 355-359.
(18) Saraiva, M. J. M.; Costa, P. P.; Goodman, D. S. J. Lab. Clin. Med.
1983, 102, 590-603.
357.
(6) Wojtczak, A.; Cody, V.; Luft, J. R.; Pangborn, W. Acta Crystallogr.
1996, D52, 758-765.
(19) Saraiva, M. J. M.; Costa, P. P.; Goodman, D. S. AdV. Neurol. 1988,
48, 189-200.
(7) Colon, W.; Kelly, J. W. Biochemistry 1992, 31, 8654-8660.
(8) McCutchen, S.; Colon, W.; Kelly, J. W. Biochemistry 1993, 32,
12119-12127.
(20) Sousa, A.; Coelho, T.; Barros, J.; Sequeiros, J. Am. J. Med. Genet.
1995, 60, 512-521.
(9) McCutchen, S. L.; Lai, Z.; Miroy, G.; Kelly, J. W.; Colon, W.
Biochemistry 1995, 34, 13527-13536.
(21) Saraiva, M. J. M. Hum. Mutat. 1995, 5, 191-196.
(22) Kelly, J. W. Curr. Op. Struct. Biol. 1996, 6, 11-17.
(23) Kelly, J. W. Structure 1997, 5, 595-600.
(10) Lai, Z.; Colon, W.; Kelly, J. W. Biochemistry 1996, 35, 6470-
6482.
(24) Kelly, J. W. Curr. Opin. Struct. Biol. 1998, 8, 101-106.
(25) Lashuel, H.; Wurth, C.; Woo, L.; Kelly, J. W. Biochemistry 1999,
38, 13560-13573.
(11) Lashuel, H. A.; Lai, Z.; Kelly, J. W. Biochemistry 1998, 37, 17851-
17864.
10.1021/ja993309v CCC: $19.00 © 2000 American Chemical Society
Published on Web 02/23/2000

Design of TTR Amyloid Fibril Inhibitors
J. Am. Chem. Soc., Vol. 122, No. 10, 2000 2179
solution turbid.11,25 The rate-determining step of TTR amyloid
fibril formation appears to be tetramer dissociation.26 Biophysi-
cal studies on WT TTR and the FAP-associated variants
demonstrate a direct relationship between tetramer stability and
the age of disease onset.8,9,11,22-24,27-29 For example, the FAP-
associated variant L55P exhibits disease onset at the earliest
age (∼20 yr) and has been demonstrated to be the least stable
TTR tetramer.8,9,11,14,25,26 In fact, the L55P TTR tetramer is the
only FAP-associated variant that dissociates to the monomeric
amyloidogenic intermediate and forms amyloid protofilaments
under physiological conditions (37 °C, pH 7.6).11,25
quantity of amyloid fibrils.33 A published 2.0 Å crystal structure
of the TTR‚(Flu)2 complex reveals that Flu nicely complements
the van der Waals surface of the two hormone binding sites in
TTR, Figure 1, A and B.33 A rearrangement in the TTR structure
induced by placement of the phenyl substituent in the inner
binding cavity and the m-CF3 group in the deepest halogen
binding pocket reorients the Ser-117 and Thr-119 side chains
from their conformations in apo TTR or the TTR‚(T4)2 struc-
tures, and in the process displaces an ordered H2O molecule.33
The ligand binding mediated TTR conformational change
reorients the Ser-117 side chains to a more interior position
where the side chains in all four subunits are close enough to
form a desolvated intersubunit hydrogen bonding network that
is thought to stabilize the TTR tetramer, Figure 1, A and B.33
The transthyretin tetramer, which possesses 222 molecular
symmetry, is constructed from identical monomers in nearly
identical conformations assembled around the central ligand
binding channel (see Figure 6 for a depiction of the tetramer
with two ligands bound to the central channel). The crystal-
lographic 2-fold axis (Z-axis, Figure 1A) running through the
channel transforms two identical dimers composing each binding
site. The other 2 perpendicular 2-fold axes (X and Y) are not
crystallographic; instead they transform subunits that are nearly
identical, only differing slightly in the conformation of the loop
regions exposed on the surface of the tetramer. The occupancy
of both ligand binding sites by Flu appears to be required to
achieve nearly complete TTR amyloid fibril inhibition with this
ligand, Table 1 and Figure 3B. It is difficult to saturate both
ligand binding sites because Flu binds with negative cooperat-
ivity, Table 2.33 In addition, many of the FAP-associated variants
(e.g., L55P) bind Flu with 2-4-fold lower affinity at both the
first and second sites relative to WT TTR (Table 2) making
amyloid fibril inhibition of these variants even more dif-
ficult.4,33,38 If 2 equiv of ligand (7.2 µM) turns out to be generally
required for complete inhibition of TTR amyloidogenesis, then
non- or positively cooperative ligand binding behavior is desired.
An ideal small molecule inhibitor would prevent TTR fibril
formation at a tetramer to ligand stoichiometry of 1:1, that is,
at a concentration of 3.6 µM.
The amyloid hypothesis states that the process of amyloid
fibril formation (involving numerous assembly intermediates)
triggers the neurodegeneration and related pathology character-
izing these diseases. The genetic evidence favoring this hy-
pothesis is strong, but unambiguous proof and a mechanistic
understanding of the pathology at the cell biology level are
lacking.14,15,30,31 A strategy in which small molecule binding is
employed to prevent the deleterious conformational changes
leading to amyloid fibril formation would be an ideal and
stringent test of this hypothesis, particularly if the small molecule
inhibited the initial misfolding step.32,33 Misfolding appears to
be the key step in amyloidogenesis in about half of the amyloid
diseases, including those based on the assembly of transthyretin,
lysozyme, and the immunoglobulin light chains.22,23,34 The key
to testing the small molecule inhibitor strategy in vivo is to
develop a compound that exhibits high selectivity and binding
affinity for TTR in plasma. Transthyretin’s natural ligand,
thyroid hormone, is not a good candidate due to its biological
activity and high affinity for thyroid binding globulin (Kd 63
pM). However, studies employing WT TTR (3.6 µM) under
acidic (pH 4.4) amyloid-forming conditions demonstrate that
T4 (10.8 µM) stabilizes >95% of the TTR tetramer against
amyloid fibril formation in vitro.32 Two molecules of T4 bind
TTR with negative cooperativity, facilitating intersubunit con-
tacts that lower the free energy of the TTR‚(T4)2 complex (Kd1
) 44 nM, Kd2 ) 2.08 µM). Stabilization of the tetramer by T4
shifts the equilibrium toward tetramer under amyloid forming
conditions and increases the activation barrier for tetramer
dissociation to the alternatively folded monomeric amyloidogen-
ic intermediate, thus inhibiting TTR amyloid fibril formation.32
The anthranilic acid substructure of Flu occupies only about
50% of the outer TTR thyroid hormone binding cavity in each
C2z symmetry equivalent binding mode, Figure 1, A and B.
Hence, an inhibitor that simultaneously occupies both sites
should bind TTR with higher affinity and selectivity.33 The
substituted phenoxazines envisioned by the structure-based
design approach outlined within prove to be excellent TTR
amyloid fibril inhibitors, especially against the most pathogenic
FAP-associated variant L55P. The approach used here is very
similar to that in one of the first manuscripts on structure-based
ligand design by Blaney and Blake, who used the TTR‚(T4)2
structure to design R-naphthyl thyroid hormone analogues.39
Previously published screening efforts identified several
compounds as potent TTR amyloid fibril inhibitors including
flufenamic acid (Flu).33,35-37 This anthranilic acid stabilizes the
tetramer and prevents WT, V30M, and L55P TTR amyloid fibril
formation at pH 4.4, a pH that normally yields the maximal
(26) Lai, Z.; McCulloch, J.; Kelly, J. W. Biochemistry 1997, 36, 10230-
10239.
(27) Nettleton, E. J.; Sunde, M.; Lai, Z.; Kelly, J. W.; Dobson, C. M.;
Robinson, C. V. J. Mol. Biol. 1998, 281, 553-564.
(28) Alves, I. L.; Altland, K.; Almeida, M. R.; Winter, P.; Saraiva, M.
J. Hum. Mutat. 1997, 9, 226-233.
(29) Quintas, A.; Saraiva, M. J. M.; Brito, R. M. M. FEBS Lett. 1997,
418, 297-300.
Experimental Section
General Methods. Unless otherwise stated reagents were purchased
from commercial suppliers and used without further purification. All
reactions were carried out under an argon atmosphere in flame-dried
glassware. Reaction progress was monitored by thin-layer chromatog-
raphy on silica gel 60 F254 coated aluminum plates (EM sciences). All
flash chromatography was performed using 230-400 mesh silica gel
60, (EM Sciences). NMR spectra were recorded either at 300, 400,
(30) Selkoe, D. J. J. Biol. Chem. 1996, 271, 18295-18298.
(31) Selkoe, D. J. Science 1997, 275, 630-631.
(32) Miroy, G. J.; Lai, Z.; Lashuel, H.; Peterson, S. A.; Strang, C.; Kelly,
J. W. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 15051-15056.
(33) Peterson, S. A.; Klabunde, T.; Lashuel, H. A.; Purkey, H.;
Sacchettini, J. C.; Kelly, J. W. Proc. Natl. Acad. Sci. U.S.A. 1998, 95,
13407-13412.
(34) Jarrett, J.; Lansbury, P. T. Cell 1993, 73, 1055-1058.
(35) Baures, P. W.; Peterson, S. A.; Kelly, J. W. Bioorg. Med. Chem.
1998, 6, 1389-1401.
(38) Almeida, M., R. Saraiva, Maria J. Eur. J. Endocrinol. 1996,135,
226-230.
(39) Blaney, J. M.; Jorgensen, E. C.; Connolly, M. L.; Ferrin, T. E.;
Langridge, R.; Oatley, S. J.; Burridge, J. M.; Blake, C. C. F. J. Med. Chem.
1982, 25, 785-790.
(36) Baures, P. W.; Oza, V. H.; Peterson, S. A.; Kelly, J. W. Bioorg.
Med. Chem. 1999, 7, 1339-1347.
(37) Oza, V. B.; Petrassi, H. M.; Purkey, H. E.; Kelly, J. W. Bioorg.
Med. Chem. Lett. 1999, 9, 1-6.

2180 J. Am. Chem. Soc., Vol. 122, No. 10, 2000
Petrassi et al.
Figure 1. (A) A simplified representation of one of the two C2x symmetrical TTR ligand binding sites. The crystallographic 2-fold axis (Z-axis)
running through the center of the ligand binding channel transforms two identical dimers composing the tetramer, each set forming a ligand binding
site (one set shown). The occupancy of one site by Flu is depicted in both of its C2z symmetry equivalent binding modes, one in blue, one in
maroon.33 For a view of both C2x symmetrical TTR ligand binding sites bound to 4 see Figure 6. (B) A framework model representing both C2z
symmetry equivalent binding modes of Flu along with their collective electron density (same color scheme as in A) generated by the 2 |Fo| - |Fc|
Φ
calc map with a σ of 2 at 2.0 Å resolution (shown in green) suggest that a tricyclic ring system would nicely occupy all of the conformational space
traced out by Flu in the outer binding cavity. (C) A simplified representation of the anticipated binding mode of the N10-phenyl phenoxazine-4,6-
dicarboxylate inhibitors conceived of by the structure-based ligand design approach described within. The N10-phenyl ring projects into the inner
binding cavity whereas the phenoxazine ring system occupies the outer cavity. (D) Framework structural model representing the solid-state structure
of 22 (the diester of 4) derived from the X-ray crystal coordinates (Supporting Information Table 1).
500, or 600 MHz on Bruker spectrometers. In the spectral data, t* refers
to an apparent triplet where the separation between the center resonance
and the left and the right wings are slightly different, suggesting second-
order contributions; the J values given refer to those separations.
Chemical shifts are reported in parts per million downfield from the
internal standard (Me4Si, 0.0 ppm). Reverse phase high performance
liquid chromatography (HPLC) was carried out on a Waters 600E multi-
solvent delivery system employing a Waters 486 tunable absorbance
detector (λdet ) 323 nm). A C18 Vydac column was employed for
preparative work (model 218TP1022 300-Å, 5-µ pore size, 22 mm i.d.
by 250 mm) and for ananlytical studies (model 218TP54 300-Å, 5-µ
pore size, 4.6 mm i.d. by 250 mm). A gradient of CH3CN was used to
elute the phenoxazine-based compounds off the reverse phase column.
A linear increase in the B solvent (95% CH3CN, 4.8% H2O, 0.2% TFA)
relative to solvent A (95% H2O, 4.8% CH3CN, 0. 2% TFA) at a flow
rate of 10 mL/min for preparative and 1.0 mL/min for analytical
columns was employed.
added to the flask by cannulation, and the flask was cooled to 0 °C
O bath). Solid NaH (3.6 g, 90 mmol, 1.5 equiv) was added by
(ice/H2
momentarily breaking the connection to the flask and condenser under
a positive argon flow. The ice/H2O bath was replaced with an oil bath,
and the reaction was heated to reflux for 1 h (upon heating the reaction
mixture, a large evolution of gas was observed). The oil bath was
removed, and the flask was allowed to cool to room temperature. A
solution of tert-butyldimethylsilyl chloride (1 M solution in THF, 90
mL, 90 mmol, 1.5 equiv) was cannulated through the septum. An oil
bath was then used to reflux the reaction for 6 h. After removal of the
oil bath the reaction mixture was poured into 1 L of a distilled ice/
H2O slurry in a 3 L separatory funnel that was extracted with EtOAc
(4 × 400 mL). The organic layers were combined and dried with
MgSO4. The solvent was then removed under reduced pressure to afford
a light-brown solid residue. The crude product was purified by flash
chromatography (900 mg of SiO2, 5% EtOAc in hexanes, Rf ) 0.84)
1
to afford 13.69 g (77%) of 1 as a white solid; H NMR (500 MHz,
CDCl3) δ 0.35 (s, 6H), 1.06 (s, 9H), 6.94-7.01 (m, 8H); 13C NMR
(125 MHz, CDCl3) δ -2.57, 20.12, 27.54, 116.03, 121.93, 123.06, 123.
22, 136.90, 151.44; FABMS (NBA/NaI) m/z 297 (M+ requires C16H23-
NOSi requires 297).
10-(tert-Butyldimethylsilyl)-phenoxazine (1). Compound 1 was
prepared by using the modified procedure of Antonio et al.40 A 1 L
flame-dried round-bottom flask equipped with a septum-capped reflux
condenser was charged with a dry stir bar and phenoxazine (10.93 g,
60 mmol) and was evacuated (high vacuum). The flask was flame-
dried and then back-filled with argon. Anhydrous THF (600 mL) was
10-(tert-Butyldimethylsilyl)-phenoxazine-4,6-dimethyldicarboxy-
late (2). Compound 2 was prepared by using the modified procedure
of Antonio et al.40 A 1 L flame-dried round-bottom flask was equipped
with a stirbar, charged with 1 (17.6 g, 59.1 mmol), septum-capped,
and evacuated. The flask was flame-dried and back-filled with argon.
(40) Antonio, Y.; Barrera, P.; Contreras, O.; Franco, F.; Galeazzi, E.;
Garcia, J.; Greenhouse, R.; Guzman, A.; Velarde, E. J. Org. Chem. 1989,
54, 2159-2165.

Design of TTR Amyloid Fibril Inhibitors
J. Am. Chem. Soc., Vol. 122, No. 10, 2000 2181
Table 1: Structural Summary of the N10-phenyl-4,6-phenoxazine Dicarboxylic Acid Based Amyloid Fibril Formation Inhibitorsa
a Accompanying the structures are the corresponding yields of fibrils formed from WT and L55P TTR at pH 4.4 in the presence of 3.6 or 7.2
µM inhibitor. When inhibitors are compared, percent conversion of tetrameric TTR into amyloid (oligomers, protofilaments, and filaments) over
a 72 h period is reported relative to TTR lacking inhibitor (100%). Hence, 100% yield corresponds to no inhibition, and 0% corresponds to complete
inhibition.
Table 2: Isothermal Titration Calorimetry Analysis of the Binding of Flu and 4 to WT and L55P TTR
(A) Dissociation Constants Kd Associated with the Binding of Inhibitors to Transthyretina
WT Kd1
WT Kd2
L55P Kd1
L55P Kd2
Flu
4
30 ( 13
79 ( 18
255 ( 97
237 ( 26
74 ( 16
357 ( 30
682 ( 80
1050 ( 80
(B) Thermodynamic Binding Parameters Associated with Inhibitor Binding to WT Transthyretinb
∆G1
∆H1
T∆S1
∆G2
∆H2
T∆S2
Flu
4
-10.25 ( 0.25
-9.67 ( 0.14
-11.4 ( 0.7
-4.52 ( 0.4
-1.15 ( 0.75
5.15 ( 0.45
-8.99 ( 0.22
-9.03 ( 0.06
-13.4 ( 0.7
-5.18 ( 0.2
-4.14 ( .73
3.85 ( .24
(C) Thermodynamic Binding Parameters Associated with Inhibitor Binding to L55P Transthyretinb
∆G1
∆H1
T∆S1
∆G2
∆H2
T∆S2
Flu
4
-9.72 ( 0.2
-8.79 ( 0.1
-11.2 ( 0.2
-6.3 ( 0.3
-1.48 ( 0.16
2.49 ( 0.33
-8.40 ( 0.12
-8.14 ( 0.05
-13.4 ( 0.3
-6.89 ( 0.3
-5.0 ( 0.33
1.25 ( 0.33
a Dissociation constants in nm units. b Binding free energies, enthalpies, and binding entropies resulting from an isothermal titration calorimetry
analysis of Flu and 4 binding to WT (B) and L55P (C) TTR (∆G, ∆H, and T∆S values are reported in units of kcal/mol).
Anhydrous THF (626 mL) was cannulated through the septum, and
the flask was cooled to 0 °C with an ice/H2O bath. A solution of n-BuLi
in hexanes (1.4 M, 140 mL, 200 mmol, 3.3 equiv) was cannulated
through the septum. After addition, the cooling bath was removed, and
the reaction was allowed to warm to room temperature and stir for 3
h. The flask was cooled to -78 °C with an acetone/dry ice slurry, and
after 10 min gaseous CO2 was added to the reaction solution via a
submerged syringe needle at a pressure of 15 psi. The cooling bath
was removed, and the reaction was allowed to stir and warm to room
temperature over the course of 1 h. The reaction mixture was poured
into a 2 L beaker containing an ice/H2O slurry (500 mL). The solution
pH was increased to 9 via the slow addition of 0.05 M KOH. The

2182 J. Am. Chem. Soc., Vol. 122, No. 10, 2000
Petrassi et al.
solution was transferred to a 3 L separatory funnel and extracted with
EtOAc (3 × 300 mL) to remove unreacted 1. The aqueous solution
was transferred to a beaker and cooled to 0 °C with an ice/H2O bath.
The solution was acidified to pH 2 with 0.5 M HCl precipitating a
white solid. The aqueous suspension (pH 2) was transferred back into
the 3 L separatory funnel, extracted with EtOAc (5 × 400 mL), and
dried with MgSO4. The organic layer was concentrated under reduced
pressure in a 500 mL round-bottom flask to afford the crude diacid, an
oil. The 500 mL flask containing the crude diacid was equipped with
a stir bar, septum-capped and evacuated. The flask was flame-dried
and then back-filled with argon. Anhydrous MeOH (160 mL) and ACS
reagent grade benzene (40 mL) were added via cannulation. Trimeth-
ylsilyldiazomethane (2 M solution in hexanes, 148 mmol, 74 mL, 2.5
equiv) was added slowly via syringe through the septum. The reaction
was then allowed to stir for 10 min and the solvent was removed under
reduced pressure. The crude product was purified by flash chromatog-
raphy (900 mg of SiO2, 40% EtOAc in hexanes, Rf ) 0.63) to afford
10-(m-Isopropylphenyl)-phenoxazine-4,6-dimethyldicarboxy-
late (22). A mixture of 3-isopropylphenyl triflouromethanesulfonate
37 (400 µL, 1.6 mmol) and 3 (400 mg, 1.33 mmol) were subjected to
the general coupling procedure outlined above. The crude product was
purified by flash chromatography (150 mg of SiO2, 20% EtOAc in
hexanes, Rf ) 0.39) to afford 252 mg (45%) of 22 as a bright yellow
1
solid. H NMR (600 MHz, CDCl3) δ 1.28 (d, J ) 7.0 Hz, 6H), 2.97
(heptet, J ) 7.0 Hz, 1H), 3.91 (s, 6H), 6.00 (d, J ) 7.9 Hz, 2H), 6.60
(t, J ) 7.9 Hz, 2H), 7.05 (d, J ) 7.5 Hz, 2H), 7.08 (d, J ) 7.9 Hz,
1H), 7.14 (s, 1H), 7.35 (d, J ) 7.5 Hz, 2H), 7.51 (t, J ) 7.9 Hz, 1H);
13C NMR (150 MHz, CDCl3) δ 23.78, 33.85, 52.02, 115.98, 119.56,
122.45, 122.93, 126.92, 127.34, 128.07, 131.09, 134.75, 138.24, 142.79,
152.62, 165.85; FABHRMS (NBA/CsI) m/z 417.1563 (M.+, C25H23-
NO5 requires 417.1576).
10-(3,5-Bis-trifluoromethylphenyl)-phenoxazine-4,6-dimethyldi-
carboxylate (23). A mixture of 3,5 bis-trifluoromethyl bromobenzene
(470 mg, 329 µL, 1.6 mmol) and 3 (400 mg, 1.33 mmol) was subjected
to the general coupling procedure outlined above. The crude product
was purified by flash chromatography (150 mg of SiO2, 20% EtOAc
in hexanes, Rf ) 0.34) to afford 132 mg (26%) of 23 as a bright yellow
1
19.4 g (79%) of 2 as a yellow solid; H NMR (500 MHz, CDCl3) δ
0.17 (s, 6H), 1.02 (s, 9H), 3.94 (s, 6H), 6.93-7.01 (m, 4H), 7.37 (d, J
) 7.7 Hz, 2H); 13C NMR (125 MHz, CDCl3) δ -3.27, 19.76, 27.63,
52.09, 120.47, 122.86, 124.89, 125.02, 137.49, 150.25, 165.80; FABMS
(NBA/NaI) m/z 413 (M+ requires C20H27NOSi requires 413).
1
solid. H NMR (400 MHz, CDCl3) δ 3.96 (s, 6H), 5.96 (dd, J ) 7.9,
1.4 Hz, 2H), 6.71 (t, J ) 7.9 Hz, 2H), 7.18 (dd, J ) 7.9, 1.4 Hz, 2H),
7.86 (s, 2H), 8.04 (s, 1H) 13C NMR (100 MHz, CDCl3) δ 52.26, 116.07,
120.44, 122.90, 123.32, 123.87, 124.06, 131.64, 133.54, 134.98, 135.32,
140.71, 142.96, 165.52; FABHRMS (NBA/CsI) m/z 643.9889 ((M +
Cs)+, C24H15F6NO5Cs requires 643.9909).
Phenoxazine-4,6-dimethyldicarboxylate (3). Compound 3 was
prepared by using the modified procedure of Antonio et al.40 A 500
mL round-bottom flask equipped with a stir bar was charged with 2
(17.9 g, 43 mmol), septum-capped, and evacuated. The flask was flame-
dried and back-filled with argon. Anhydrous THF (100 mL) was
cannulated through the septum. Tetrabutylammonium fluoride (1 M in
THF, 65 mL, 65 mmol, 1.5 equiv) was added to the reaction via syringe.
The reaction was stirred for 1 h at room temperature and then poured
into 300 mL of H2O in a 2 L separatory funnel. Compound 3 was
extracted with EtOAc (3 × 200 mL). The organic layers were combined
and dried with MgSO4, and the solvent was removed under reduced
pressure. The product residue was recrystallized from hexanes:EtOAc
10-(p-Methylbenzoate)-phenoxazine-4,6-dimethyldicarboxylate (24).
A mixture of methyl 4-bromobenzoate (346 mg, 1.6 mmol) and 3 (400
mg, 1.33 mmol) were subjected to the general coupling procedure
outlined above. The crude product was purified by flash chromatography
(150 mg of SiO2, 20% EtOAc in hexanes, Rf ) 0.35) to afford 255 mg
1
(44%) of 24 as a bright yellow solid; H NMR (500 MHz, CDCl3) δ
3.91 (s, 6H), 3.98 (s, 3H), 6.00 (d, J ) 8.1 Hz, 2H), 6.63 (t*, J ) 7.7,
8.1 Hz, 2H), 7.10 (d, J ) 7.7 Hz, 2H), 7.41 (d, J ) 8.5 Hz, 2H), 8.29
(d, J ) 8.5 Hz, 2H); 13C NMR (125 MHz, CDCl3) δ 51.85, 52.10,
112.96, 115.85, 119.61, 122.84, 130.32, 132.40, 133.74, 142.48, 142.54,
161.06, 165.36, 165.62; FABHRMS (NBA/CsI) m/z 434.1229 (M+H+,
C24H20NO7 requires 434.1240).
10-(m-Ethylphenyl)-phenoxazine-4,6-dimethyldicarboxylate (25).
A mixture of 3-ethylphenyl triflouromethanesulfonate 36 (400 µL, 1.6
mmol) and 3 (400 mg, 1.33 mmol) was subjected to the general coupling
procedure outlined above. The crude product was purified by flash
chromatography (150 mg of SiO2, 20% EtOAc in hexanes, Rf ) 0.31)
to afford 401 mg (74%) of 25 as a bright yellow solid; 1H NMR (500
MHz, CDCl3) δ 1.24 (t, J ) 7.4 Hz, 3H), 2.69 (q, J ) 7.4 Hz, 2H),
3.93 (s, 6H), 6.00 (d, J ) 8.1 Hz, 2H), 6.56 (t*, J ) 8.1, 7.9 Hz, 2H),
7.02-7.10 (m, 4H), 7.29 (d, J ) 7.4 Hz, 1H), 7.48 (t, J ) 7.7 Hz,
1H); 13C NMR (125 MHz,CDCl3) δ 15.23, 28.49, 52.00, 116.02, 119.54,
122.46, 122.90, 127.25, 128.35, 129.39, 131.07, 134.69, 138.24, 142.76,
147.86, 165.82; FABHRMS (NBA/CsI) m/z 403.1420 (M+, C24H21-
NO5 requires 403.1420).
1
(9:1) to afford 11.24 g (87%) of 3 as a yellow crystalline solid. H
NMR (600 MHz, DMSO-d6) δ 3.80 (s, 6H), 6.63 (dd, J ) 1.7, 7.9 Hz,
2H), 6.83 (t, J ) 7.9 Hz, 2H), 6.89 (dd, J ) 1.7, 7.9 Hz, 2H), 8.64, (s,
1H); 13C NMR (150 MHz, DMSO d6) δ 52.03, 116.15, 119.49, 121.08,
124.03, 132.98, 141.21, 165.46; FABMS (NBA/NaI) m/z 299 (M+
requires C25H23NO5 requires 299).
General Procedure for the Synthesis of N-Phenyl Phenoxazines.
The aryl coupling procedure utilized was that reported by Buchwald
and Hartwig.41,42 A flame-dried 10 mm × 13 cm borosilicate tube,
equipped with a stirbar and septum-capped, was charged with 3 (400
mg, 1.3 mmol), palladium dibenzylidene acetone, Pd2(dba)3 (61 mg,
0.06 mmol, 0.05 equiv), (()-Binap (62 mg, 0.08 mmol, 0.075 equiv),
Cs2CO3 (645 mg, 1.82 mmol, 1.4 equiv), and the aryl halide or triflate
(1.6 mmol, 1.2 equiv). In the case of nonvolatile reagents, the reaction
tube was evacuated for 1 h and back-filled with argon. In the case of
a volatile reagent the flask was simply purged with argon. In both cases,
anhydrous toluene (2.4 mL) was added through the septum, and the
reaction mixture was heated to 100 °C for 36 h in an oil bath. The
reaction mixture was filtered through a cellulose paper filter, and the
solid residue was washed with CH2Cl2 (3 × 10 mL). The solvent was
removed under reduced pressure, and the resulting crude compounds
were purified by flash chromatography.
10-(m-Trifluoromethylphenyl)-phenoxazine-4,6-dimethyldicar-
boxylate (21). A mixture of 3-bromobenzotrifluoride (400 µL, 1.6
mmol) and 3 (400 mg, 1.33 mmol) was subjected to the general coupling
procedure outlined above. The crude product was purified by flash
chromatography (150 mg of SiO2 20% EtOAc in hexanes, Rf ) 0.35)
to afford 118 mg (26%) of 21 as a bright-yellow solid; 1H NMR (300
MHz, CDCl3) δ 3.94 (s, 6H), 5.96 (d, J ) 7.9 Hz, 2H), 6.66 (m, 2H),
7.11 (d, J ) 7.9 Hz, 2H), 7.11 (d, J ) 7.9 Hz, 2H), 7.50-7.83 (m,
4H); 13C NMR (75 MHz, CDCl3) δ 52.13, 115.93, 119.98, 123.24,
125.06, 125.87, 132.16, 134.07, 134.39, 142.76, 165.67; MALDIHRMS
(DHB) m/z 446.0874 ((M + Na)+, C23H16F3NO5Na requires 446.0878).
10-(m-tert-Butylphenyl)-phenoxazine-4,6-dimethyldicarboxy-
late (26). A mixture of 3-tert-butylphenyl triflouromethanesulfonate
(400 µL, 1.6 mmol) and 3 (400 mg, 1.33 mmol) was subjected to the
general coupling procedure outlined above. The crude product was
purified by flash chromatography (150 mg of SiO2, 20% EtOAc in
hexanes, Rf ) 0.31) to afford 285 mg (49%) of 26 as a bright-yellow
1
solid. H NMR (400 MHz, CDCl3) δ 1.31 (s, 9H), 3.96 (s, 6H), 5.90
(dd, J ) 8.1, 1.6 Hz, 2H), 6.52 (t*, J ) 8.1, 7.8 Hz, 2H), 6.95-6.97
(m, 2H), 6.97-7.01 (m, 1H), 7.18-7.20 (m, 1H), 7.42 (m, 2H); 13C
NMR (75 MHz, CDCl3) δ 31.16, 34.83, 52.04, 116.00, 119.54, 122.46,
122.95, 125.76, 127.00, 130.84, 134.83, 138.03, 142.80, 155.04, 165.87;
FABHRMS (NBA/CsI) m/z 454.1644 ((M + Na)+, C26H25NO5Na
requires 454.1630).
10-(p-Methyphenyl)-phenoxazine-4,6-dimethyldicarboxylate (27).
A mixture of 4-bromotoluene (274 mg, 197 µL, 1.6 mmol) and 3 (400
mg, 1.33 mmol) was subjected to the general coupling procedure
outlined above. The crude product was purified by flash chromatog-
raphy (150 mg of SiO2, 30% EtOAc in hexanes, Rf ) 0.48) to afford
362 mg (70%) of 27 as a bright yellow solid; 1H NMR (500
MHz, CDCl3) δ 2.45 (s, 3H), 3.93 (s, 6H), 6.00 (dd, J ) 8.1, 1.5 Hz,
(41) Wolfe, J. P.; Buchwald, S. L. Tetrahedron Lett. 1997, 38, 6359-
6362.
(42) Louie, J.; Driver, M. S.; Hamann, B. C.; Hartwig, J. F. J. Org. Chem.
1997, 62, 1268-1273.

Design of TTR Amyloid Fibril Inhibitors
J. Am. Chem. Soc., Vol. 122, No. 10, 2000 2183
2H), 6.60 (t*, J ) 8.1, 7.7 Hz, 2H), 7.04 (dd, J ) 7.7, 1.5 Hz,
2H), 7.16 (d, J ) 8.1 Hz, 2H), 7.39 (d, J ) 8.1 Hz, 2H); 13C NMR
(75 MHz,CDCl3) δ 21.26, 52.16, 116.11, 119.67, 122.57, 123.02,
130.13, 132.01, 134.72, 135.68, 138.98, 142.92, 166.03; FABHRMS
(NBA/CsI) m/z 412.1161 ((M + Na)+, C23H19NO5Na requires
412.1146).
10-(p-Nitrophenyl)-phenoxazine-4,6-dimethyldicarboxylate (28).
A mixture of 4-nitrophenyl trifluoromethanesulfonate (434 mg, 1.6
mmol) and 3 (400 mg, 1.33 mmol) was subjected to the general coupling
procedure outlined above. The crude product was purified by flash
chromatography (150 mg of SiO2, 20% EtOAc in hexanes, Rf )0.37)
to afford 496 mg (99%) of 28 as a bright yellow solid; 1H NMR (500
MHz, CDCl3) δ 3.94 (s, 6H), 6.10 (dd, J ) 7.7, 1.5 Hz, 2H), 6.69 (t,
J ) 8.1 Hz, 2H), 7.16 (dd, J ) 7.7, 1.5 Hz, 2H), 7.55 (d, J ) 8.8 Hz,
2H), 8.47 (d, J ) 8.8 Hz, 2H); 13C NMR (125 MHz, CDCl3) δ 52.16,
116.59, 120.20, 123.14, 123.79, 126.60, 131.09, 133.44, 143.15, 144.93,
147.20, 165.46; FABHRMS (NBA/CsI) m/z 421.1036 ((M + H)+,
C22H17N2O7 requires 421.1048).
10-(p-Cyanophenyl)-phenoxazine-4,6-dimethyldicarboxylate (29).
A mixture of 4-bromobenzonitrile (291 mg, 1.6 mmol) and 3 (400 mg,
1.33 mmol) was subjected to the general coupling procedure outlined
above. The crude product was purified by flash chromatography (150
mg of SiO2, 30% EtOAc in hexanes, Rf ) 0.33) to afford 110 mg (20%)
of 29 as a bright-yellow solid; 1H NMR (500 MHz, CDCl3) δ 3.93 (s,
6H), 6.02 (d, J ) 8.1 Hz, 2H), 6.67 (t*, J ) 7.7, 8.1 Hz, 2H), 7.14 (d,
J ) 8.1 Hz, 2H), 7.49 (d, J ) 8.1 Hz, 2H), 7.93 (d, J ) 8.1 Hz, 2H);
13C NMR (125 MHz, CDCl3) δ 52.21, 112.76, 116.26, 117.75, 120.23,
123.14, 123.63, 131.52, 133.61, 135.23, 142.96, 143.11, 165.58;
FABHRMS (NBA/CsI) m/z 400.1069 (M+, C23H16N2O5 requires
400.1059).
10-(Phenyl)-phenoxazine-4,6-dimethyldicarboxylate (33). A mix-
ture of bromobenzene (350 µL, 1.6 mmol) and 3 (400 mg, 1.33 mmol)
was subjected to the general coupling procedure outlined above. The
crude product was purified by flash chromatography (150 mg of SiO2,
20% EtOAc in hexanes, Rf ) 0.29) to afford 223 mg (48%) of 33 as
a bright-yellow solid; 1H NMR (500 MHz, CDCl3) δ 3.81 (s, 6H), 5.88
(dd, J ) 8.1, 1.5 Hz, 2H), 6.50 (t*, J ) 8.1, 7.7 Hz, 2H), 6.95 (dd, J
) 7.7, 1.5 Hz, 2H), 7.18 (m, 2H), 7.39 (m, 1H), 7.50 (m, 2H); 13C
NMR (125 MHz, CDCl3) δ 52.00, 115.96, 119.60, 122.55, 122.91,
128.84, 130.32, 131.26, 134.60, 138.29, 142.72. 165.76; FABHRMS
(NBA/NaI) m/z 375.1107 (M+, C22H17NO5 requires 375.1099).
10-(m-Trifluoromethylpyrimidine)-phenoxazine-4,6-dimethyldi-
carboxylate (34). A mixture of 2 chloro-4-trifluoromethyl pyrimidine
(657 mg, 434 µL, 3.2 mmol) and 3 (400 mg, 1.33 mmol) was subjected
to the general coupling procedure outlined above. The crude product
was purified by chromatography (150 mg of SiO2, 20% EtOAc in
hexanes, Rf ) 0.33) to afford 239 mg (41%) of 34 as a slightly yellowish
solid; 1H NMR (500 MHz, CDCl3) δ 4.01 (s, 6H), 7.18 (d, J ) 4.8 Hz,
1H), 7.23 (t*, J ) 7.7, 8.1 Hz, 2H), 7.68 (dd, J ) 7.7, 1.5 Hz, 2H),
7.94 (dd, J ) 8.1, 1.5 Hz, 2H), 8.63 (d, J ) 4.78 Hz, 1H); 13C NMR
(125 MHz, CDCl3) δ 52.31, 110.09, 121.25, 122.67, 128.14, 130.23,
150.39, 159.13, 160.28, 165.35; FABHRMS (NBA/NaI) m/z 445.0876
(M+, C22H17NO5 requires 445.0886).
General Procedure for Ester Hydrolysis. Basic saponification was
performed by dissolving methyl esters 21-35 in a THF:MeOH:H2O
(3:1:1) ternary solvent system in a round-bottom flask equipped with
a stir bar. To each 1 equiv of methyl ester was added 4 equiv of LiOH,
and the reaction was allowed to stir until complete (typically 4 h) as
determined by TLC or analytical reverse phase HPLC monitoring. The
reaction mixtures were poured into a separatory funnel and acidified
to pH 2 with 1 M HCl (pH paper). The diacids were extracted with
EtOAc, CHCl3, or a mixture of thereof. In cases where purification
was necessary, compounds were purified either by recrystallization
(EtOAc/hexanes), flash chromatography, or by reverse phase HPLC.
10-(m-Trifluoromethylphenyl)-phenoxazine-4,6-dicarboxylate (4).
Dimethyl ester 21 (303 mg, 0.683 mmol) was subjected to the ester
hydrolysis procedure outlined above. The crude product was dissolved
in DMSO and purified by reverse phase HPLC, employing a linear
10-(p-Methoxyphenyl)-phenoxazine-4,6-dimethyldicarboxylate (30).
A mixture of 4-iodoanisole (278 mg, 1.2 mmol), palladium (II) acetate
(7 mg, 0.025 mmol), Cs2CO3 (456 mg, 1.4 mmol), dioxane (2 mL),
(+)-(S)-1-[(R)-2-(diphenylphosphino)ferrocenyl] ethyl methyl ether (50
mg, 0.05 mmol), and 3 (298 mg, 1.0 mmol) was charged into a flame-
dried 10 mL round-bottom flask equipped with a stir bar and a septum-
capped reflux condenser. The reaction mixture was heated for 24 h at
reflux with an oil bath and then worked up according to the general
procedure. The crude product was purified by flash chromatography
(150 mg of SiO2, 30% EtOAc in hexanes, Rf ) 0.44) to afford 110 mg
1
gradient of 20-80% B over 25 min to yield 277 mg of 4 (98%); H
NMR (600 MHz DMSO-d6) δ 5.98 (dd, J ) 7.9, 1.5 Hz, 2H), 6.78 (t,
J ) 7.9 Hz, 2H), 7.11 (dd, J ) 7.9, 1.5 Hz, 2H), 7.82-7.94 (m, 1H),
7.97-8.03 (m, 3H); 13C NMR (150 MHz, DMSO-d6) δ 117.10, 120.51,
123.79, 124.74, 126.87, 128.11, 134.77, 136.27, 142.68, 166.52;
FABHRMS (NBA/NaI) m/z 415.0668 (M+ requires C25H23NO5 requires
415.0678).
1
(20%) of 30 as a bright-yellow solid; H NMR (500 MHz, CDCl3) δ
3.79 (s, 3H), 3.88 (s, 6H), 5.92 (dd, J ) 8.1, 1.5 Hz, 2H), 6.61 (t, J )
8.1 Hz, 2H), 6.95 (dd, J ) 8.1, 1.5 Hz, 2H), 7.00 (d, J ) 8.8 Hz, 2H),
7.10 (d, J ) 8.8 Hz, 2H); 13C NMR (125 MHz,CDCl3) δ 52.04, 55.43,
116.00, 116.40, 119.57, 122.46, 122.95, 130.37, 131.38, 135.00, 142.82,
159.53, 165.89; FABHRMS (NBA/CsI) m/z 428.1128 (M + Na+,
C23H19NO6Na requires 428.1110).
10-(m-Methylphenyl)-phenoxazine-4,6-dimethyldicarboxylate (31).
A mixture of 3-methylphenyl trifluoromethanesulfonate 35 (400 µL,
1.6 mmol) and 3 (400 mg, 1.33 mmol) was subjected to the general
coupling procedure outlined above. The crude product was purified by
flash chromatography (150 mg of SiO2, 20% EtOAc in hexanes, Rf )
0.31) to afford 319 mg (63%) of 31 as a bright yellow solid; 1H NMR
(300 MHz, CDCl3) δ 2.41 (s, 3H), 3.96 (s, 6H), 6.00 (d, J ) 7.9 Hz,
2H), 6.60 (t, J ) 7.9 Hz, 2H), 7.02-7.10(m, 4H), 7.29 (d, J ) 7.5 Hz,
1H), 7.48 (m, 1H); 13C NMR (75 MHz, CDCl3) δ 21.23, 52.02, 116.02,
119.51, 122.46, 122.89, 127.08, 129.59, 130.98, 131.27, 134.65, 138.14,
141.49, 142.74, 165.81; FABHRMS (NBA/NaI) m/z 412.1146 (M+,
C23H19NO5Na requires 412.1161).
10-(p-trifluoromethylphenyl)-phenoxazine-4,6-dimethyldicarbox-
ylate (32). A mixture of 4-iodobenzotrifluoride (435 mg, 235 µL, 1.6
mmol) and 3 (400 mg, 1.33 mmol) was subjected to the general coupling
procedure outlined above. The crude product was purified by flash
chromatography (150 mg of SiO2, 20% EtOAc in hexanes, Rf ) 0.32)
to afford 119 mg (19%) of 32 as a bright-yellow solid; 1H NMR (600
MHz, CDCl3) δ 3.93 (s, 6H), 5.98 (dd, J ) 7.9, 1.7 Hz, 2H), 6.63 (t,
J ) 7.9 Hz, 2H), 7.10 (dd, J ) 7.9, 1.7 Hz, 2H), 7.47 (d, J ) 8.3 Hz,
2H), 7.88 (d, J ) 8.3 Hz, 2H); 13C NMR (150 MHz, CDCl3) δ 52.13,
116.02, 120.02, 123.08, 123.23, 128.51, 131.26, 133.98, 141.96, 142.78,
165.67; FABHRMS (NBA/NaI) m/z 466.0868 (M + Na+, C23H16F3-
NO5Na requires 466.0878).
10-(m-Isopropylphenyl)-phenoxazine-4,6-dicarboxylate (5). Di-
methyl ester 22 (100 mg, 0.24 mmol) was subjected to the ester
hydrolysis procedure outlined above. Extraction afforded 90 mg of pure
1
5 (96%); Rf ) 0.38 (0.1% TFA, 39.9% EtOAc in hexanes) H NMR
(600 MHz, DMSO-d6) δ 1.24 (d, J ) 7.0 Hz, 6H), 2.96 (heptet, J )
7.0 Hz, 1H), 5.97 (d, J ) 7.9 Hz, 2H), 6.74 (t, J ) 7.9 Hz, 2H), 7.09
(d, J ) 7.9 Hz, 2H), 7.21 (d, J ) 7.9 Hz, 1H), 7.28 (s, 1H), 7.43 (d,
J ) 7.9 Hz, 2H), 7.59 (t, J ) 7.9 Hz, 1H); 13C NMR (150 MHz, DMSO-
d6) δ 23.75, 33.31, 116.23, 119.38, 122.70, 123.89, 127.18, 127.32,
127.86, 131.53, 134.34, 137.81,141.93, 152.49, 165.68; MALDIFTMS
(DHB) 412.1165 m/z ((M + Na)+,C23H19NO5Na requires 412.1161).
10-(3,5-Bis-trifluoromethylphenyl)-phenoxazine-4,6-dicarboxy-
late (6). Dimethyl ester 23 (150 mg, 0.293 mmol) was subjected to
the ester hydrolysis procedure outlined above. Extraction afforded 140
mg of pure 6 (99%); Rf ) 0.41 (0.1% TFA, 39.9% EtOAc in hexanes);
1H NMR (600 MHz, DMSO-d6) δ 6.04 (d, J ) 7.9 Hz, 2H), 6.77 (t, J
) 7.9 Hz, 2H), 7.13 (d, J ) 7.5 Hz, 2H), 8.34 (s,1H), 8.38 (s, 2H); 13
C
NMR (150 MHz, DMSO-d6 ) δ 116.54, 119.85, 121.96, 123.34, 123.87,
132.78, 133.71, 140.31, 141.94, 165.67; MALDIFTMS (DHB) 483.0556
m/z (M+ requires C22H11F6NO5 requires 483.0541).
10-(p-Carboxyphenyl)-phenoxazine-4,6-dicarboxylate (7). Di-
methyl ester 24 (225 mg, 0.519 mmol) was subjected to the ester
hydrolysis procedure outlined above. Extraction afforded 200 mg of
1
pure 7 (99%); Rf ) 0.17 (0.1% TFA, 39.9% hexanes in EtOAc); H
NMR (500 MHz, DMSO-d6) δ 6.05 (dd, J ) 7.7, 1.5 Hz, 2H), 6.77
(t*, J ) 8.1, 7.7 Hz, 2H), 7.11 (dd, J ) 8.1, 1.3 Hz, 2H), 7.59 (d, J )

2184 J. Am. Chem. Soc., Vol. 122, No. 10, 2000
Petrassi et al.
8.5 Hz, 2H), 8.22 (d, J ) 8.5 Hz, 2H); 13C NMR (150 MHz, DMSO-
d6) δ 116.31, 120.09, 123.15, 123.83, 130.61, 131.50, 132.62, 133.88,
141.95, 142.04, 165.68, 166.58; MALDIFTMS (DHB) 391.0684 m/z
(M+ requires C21H12NO7 requires 391.0692).
(t*, J ) 7.9, 8.1 Hz, 2H), 7.09 (d, J ) 8.1 Hz, 2H), 7.22 (d, J ) 7.7
Hz, 1H), 7.26 (s, 1H), 7.39 (d, J ) 7.7 Hz, 1H), 7.58 (t, J ) 7.7 Hz,
1H); 13C NMR (150 MHz, DMSO-d6) δ 20.86, 116.29, 119.44, 122.64,
123.84, 126.99, 129.92, 130. 37, 131.41, 134.25, 137.78, 141.51. 141.82,
165.67; MALDIFTMS (DHB) 361.0963 m/z (M+ requires C24H21NO5
requires 361.0950).
10-(m-Ethylphenyl)-phenoxazine-4,6-dicarboxylate (8). Dimethyl
ester 25 (100 mg, 0.25 mmol) was subjected to the ester hydrolysis
procedure outlined above. Extraction afforded 93 mg of pure 8 (99%);
Rf ) 0.17 (0.1% TFA, 39.9% hexanes in EtOAc); 1H NMR (600 MHz
DMSO-d6) δ 1.22 (t, J ) 7.7 Hz, 3H), 2.70 (q, J ) 7.7 Hz, 2H), 6.00
(d, J ) 8.1 Hz, 2H), 6.76 (t, J ) 8.1 Hz, 2H), 7.10 (d, J ) 8.1 Hz,
2H), 7.28 (s, 1H), 7.24 (d, J ) 7.7 Hz, 1H), 7.42 (d, J ) 7.7 Hz 1H),
7.60 (t, J ) 7.7 Hz, 1H); 13C NMR (150 MHz, DMSO-d6) δ 15.75,
29.36, 118.23, 124.52, 125.17, 128.33, 129.52, 130.43, 132.27, 135.53,
139.45, 144.06, 149.21, 165.09; MALDIFTMS (DHB) 398.1012 m/z
((M + Na)+ requires C22H17NO5Na requires 398.1004).
10-(p-Phenol)phenoxazine-4,6 dicarboxylate (15). Dimethyl ester
30 (30 mg, 0.08 mmol) and a stir bar were added to a septum- capped
10 mL round-bottom flask. The flask was evacuated, flame-dried, and
back-filled with argon. Anhydrous CHCl2 (1.87 mL) was added to the
flask via syringe. The flask was cooled to 0 °C with an ice/H2O bath.
A solution of BBr3 (1 M, 1.11 mL, 1.11 mmol, 15 equiv) was added
via syringe. The reaction was allowed to warm to room temp by
removing the ice/H2O bath and was stirred for 24 h. The reaction was
quenched by the dropwise addition of H2O (5 mL). The pH of the
reaction mixture was increased to 10 via the addition of 1 M KOH
and acidified to pH 2 via the addition of 1 M HCl. The acidified reaction
mixture was then transferred to a 250 mL separatory funnel and
extracted with CHCl3 (3 × 50 mL). The organic layers were combined
and dried with MgSO4, and the solvent was removed under reduced
pressure. Flash chromatography of the crude product (50 mg of SiO2,
0.1% TFA, 39.9% hexanes in EtOAc, Rf ) 0.30) yields 15.9 mg of
pure 15 (60%); 1H NMR (500 MHz, DMSO-d6) δ 6.05 (d, J ) 8.1 Hz,
2H), 6.76 (t*, J ) 7.7, 8.1 Hz, 2H), 7.02 (d, J ) 7.7 Hz, 2H), 7.08 (d,
J ) 7.7 Hz, 2H), 7.20 (d, J ) 7.7 Hz, 2H); 13C NMR (125 MHz,
DMSO-d6) δ 116.27, 117.97, 119.19, 122.49, 123.80, 128.45, 131.11,
134.73, 141.88, 157.76, 165.60; MALDIFTMS (DHB) 363.0747 m/z
(M+ requires C24H21NO5 requires 363.0743).
10-(p-Trifluoromethylphenyl)phenoxazine-4,6-dicarboxylate (16).
Dimethyl ester 32 (35 mg, 0.08 mmol) was subjected to the ester
hydrolysis procedure outlined above. Extraction afforded 32 mg of pure
16 (98%). Rf ) 0.50 (0.1% TFA, 39.9% hexanes in EtOAc); 1H NMR
(500 MHz, DMSO-d6) δ 6.04 (d, J ) 7.7 Hz, 2H), 6.78 (t*, J ) 7.7,
8.1 Hz, 2H), 7.12 (d, J ) 8.1 Hz, 2H), 7.73 (d, J ) 8.8 Hz, 2H), 8.06
(d, J ) 8.8 Hz, 2H); 13C NMR (125 MHz, DMSO-d6) δ 116.38, 119.81,
123.05, 123.85, 128.74, 128.95, 129.30, 129.56, 131.59, 133.68, 141.84,
165.60; MALDIFTMS (DHB) m/z 438.0548 ((M + Na)+ requires
C21H15NO6Na requires 438.0565).
10-(m-tert-Butylphenyl)-phenoxazine-4,6-dicarboxylate (9). Di-
methyl ester 26 (100 mg, 0.2 mmol) was subjected to the ester
hydrolysis procedure outlined above. Extraction afforded 91 mg of pure
1
9 (98%); Rf ) 0.16 (0.1% TFA, 39.9% hexanes in EtOAc); H NMR
(600 MHz, DMSO-d6) δ 1.31 (s, 9H), 5.98 (d, J ) 7.9 Hz, 2H), 6.77
(t, J ) 7.9 Hz, 2H), 7.09 (d, J ) 7.9 Hz, 2H), 7.24 (d, J ) 7.0 Hz,
1H), 7.43 (s, 1H), 7.59-7.64 (m, 2H); 13C NMR (150 MHz, DMSO-
d6) δ 31.00, 34.74, 103.46, 116.14, 119.40, 122.61, 123.89, 125.95,
126.86, 126.98, 131.26, 134.41, 137.59, 141.91, 154.83, 165.67;
MALDIFTMS (DHB) 426.1325 m/z ((M + Na)+ requires C24H21NO5-
Na requires 426.1317).
10-(p-Methylphenyl)phenoxazine-4,6-dicarboxylate (10). Dimethyl
ester 27 (150 mg, 0.385 mmol) was subjected to the ester hydrolysis
procedure outlined above. Extraction afforded 129 mg of pure 10 (86%).
Rf ) 0.51 (0.1% TFA, 39.9% hexanes in EtOAc); 1H NMR (500 MHz,
DMSO-d6) δ 2.42 (s, 3H), 6.00 (dd, J ) 8.1, 1.5 Hz, 2H), 6.75 (t, J )
8.1 Hz, 2H), 7.07 (dd, J ) 8.1, 1.5 Hz, 2H), 7.31 (d, J ) 8.4 Hz, 2H),
7.50 (d, J ) 8.4 Hz, 2H); 13C NMR (150 MHz, DMSO-d6) δ 21.70,
117.08, 120.35, 123.49, 124.70, 130.78, 133.05, 135.28, 136.04, 139.71,
142.69, 166.55; MALDIFTMS (DHB) 361.0941 m/z (M+ requires
C24H21NO5 requires 361.0950).
10-(p-Nitrophenyl)phenoxazine-4,6-dicarboxylate (11). Dimethyl
ester 28 (190 mg, 0.45 mmol) was subjected to the ester hydrolysis
procedure outlined above. Extraction afforded 174 mg of pure 11 (98%).
Rf ) 0.52 (0.1% TFA, 39.9% hexanes in EtOAc); 1H NMR (600 MHz,
DMSO-d6) δ 6.16 (d, J ) 7.9 Hz, 2H), 6.80 (t, J ) 7.9 Hz, 2H), 7.15
(d, J ) 7.9 Hz, 2H), 7.78 (d, J ) 8.8 Hz, 2H), 7.50 (d, J ) 8.8 Hz,
2H); 13C NMR (150 MHz, DMSO-d6) δ 116.88, 120.24, 123.51, 123.84,
126.91, 131.55, 133.45, 142.23, 144.45, 147.15, 165.62; MALDIFTMS
(DHB) 392.0647 m/z (M+ requires C20H12N2O7 requires 392.0645).
10-(p-Cyanophenyl)phenoxazine-4,6-dicarboxylate (12). Dimethyl
ester 29 (35 mg, 0.09 mmol) was subjected to the ester hydrolysis
procedure outlined above. The crude product was dissolved in DMSO
and purified by RP-HPLC employing a linear gradient of 50-100% B
over (retention time ) 12.1 min) 25 min to yield 29 mg of 12 (89%);
1H NMR (600 MHz, DMSO-d6) δ 6.07 (dd, J ) 8.1, 1.5 Hz, 2H), 6.78
(t, J ) 8.1 Hz, 2H), 7.13 (d, J ) 8.1, 1.5 Hz, 2H), 7.72 (d, J ) 8.4 Hz,
2H), 8.18 (d, J ) 8.4 Hz, 2H); 13C NMR (150 MHz, DMSO-d6) δ
112.77,117.43, 119.14, 120.77, 124.05, 124.76, 132.58, 134.41, 136.69,
142.79, 143.40, 166.54; MALDIFTMS (DHB) 372.0753 m/z (M+
requires C21H12N2O5 requires 372.0746).
10-(Phenyl)phenoxazine-4,6-dicarboxylate (17). Dimethyl ester 33
(17.1 mg, 0.05 mmol) was subjected to the ester hydrolysis procedure
outlined above. The compound was dissolved in DMSO and purified
by reverse phase HPLC, employing a linear gradient of 40-100% B
over (retention time ) 35.1 min) 55 min to yield 15 mg of 17 (98%).
Rf ) 0.41 (0.1% TFA, 39.9% hexanes in EtOAc); 1H NMR (600 MHz,
THF-d8)δ 6.10 (d, J ) 7.9 Hz, 2H), 6.72 (t*, J ) 7.9, 8.1 Hz, 2H),
7.38-7.42 (m, 4H), 7.54 (t, J ) 7.5 Hz, 1H), 7.68 (t, J ) 7.9 Hz, 2H);
13C NMR (125 MHz, DMSO-d6) δ 116.19, 119.45, 122.68, 123.79,
129.23, 130.18, 131.63, 134.23, 137.84, 141.81, 165.59; MALDIFTMS
(DHB) 370.0690 m/z ((M + Na)+ requires C21H15NO6Na requires
370.0691).
10-(tert-butyldimethylsilyl)-phenoxazine-4,6-dicarboxylate (18).
Crude 18 whose synthesis was described above was dissolved in DMSO
and purified by reverse phase HPLC, employing a linear gradient of
40-100% B over (retention time ) 35.1 min) 55 min, Rf ) 0.41 (0.1%
1
TFA, 39.9% hexanes in EtOAc); H NMR (500 MHz, THF-d8)δ 0.29
(s, 6H), 0.94 (s, 9H), 7.09 (t, J ) 7.7 Hz, 2H), 7.19 (dd, J ) 7.4, 1.6
Hz, 2H), 7.7 (dd, J ) 7.4, 1.6 Hz, 2H); 13C NMR (125 MHz, DMSO-
d6) δ -1.87, 20.75, 28.03, 117.61, 119.57, 121.32, 122.85, 124.65,
125.14, 125.88, 126.44, 133.54, 137.87, 142.37, 150.44, 166.32, 166.54;
HRMS (NBA/CsI) m/z 386.1419 ((M + H)+ requires C20H24NO5Si
requires 386.1424).
Phenoxazine-4,6-dicarboxylate (19). The impure diacid 2 (114 mg,
0.29 mmol) (obtained before esterification with trimethylsilyl diazo-
methane) was added to a 500 mL flask with a stir bar. The flask was
septum-capped, evacuated, flame-dried, and back-filled with argon.
Anhydrous THF was cannulated through the septum. Tetrabutlyam-
monium fluoride (1 M in THF, 0.45 mL, 0.44 mmol, 1.5 equiv) was
added to the reaction via syringe. After the reaction stirred for 1 h, it
was poured into a 500 mL beaker containing 50 mL of H2O. The
contents of the beaker were transferred to a 250 mL separatory funnel
and extracted with EtOAc (3 × 500 mL). The organic layers were
10-(p-Methoxyphenyl)phenoxazine-4,6-dicarboxylate (13). Di-
methyl ester 30 (28.2 mg, 0.069 mmol) was subjected to the ester
hydrolysis procedure outlined above. Extraction afforded 25.1 mg of
1
pure 13 (98%). Rf ) 0.50 (0.1% TFA, 39.9% hexanes in EtOAc); H
NMR (600 MHz, DMSO-d6) δ 3.84 (s, 3H), 6.00 (d, J ) 7.9 Hz, 2H),
6.74 (t, J ) 7.9 Hz, 2H), 7.06 (d, J ) 7.9 Hz, 2H), 7.20 (d, J ) 8.8
Hz, 2H), 7.33 (d, J ) 8.8 Hz, 2H); 13C NMR (150 MHz, DMSO-d6) δ
55.48, 116.24, 116.76, 119.40, 122.61, 123.83, 130.10, 131.35, 134.64,
141.89, 159.35, 165.67; MALDIFTMS (DHB) 377.0893 m/z (M+
requires C21H15NO6 requires 377.0899).
10-(m-Methylphenyl)phenoxazine-4,6-dicarboxylate (14). Di-
methyl ester 31 (100 mg, 0.26 mmol) was subjected to the ester
hydrolysis procedure outlined above. Extraction afforded 91 mg of pure
14 (96%). Rf ) 0.51 (0.1% TFA, 39.9% hexanes in EtOAc); 1H NMR
(500 MHz, DMSO-d6) δ 2.40 (s, 3H), 6.01 (d, J ) 7.9 Hz, 2H), 6.76

Design of TTR Amyloid Fibril Inhibitors
J. Am. Chem. Soc., Vol. 122, No. 10, 2000 2185
combined and dried with MgSO4, and the solvent was removed under
reduced pressure. The crude product was dissolved in DMSO and
purified by reverse phase HPLC, employing a linear gradient of 40-
100% B over (retention time ) 35.1 min) 55 min to yield 45 mg of 19
of TTR suggest that a tricyclic ring system would be ideal for occupancy
of this cavity. Several computer modeling programs including the
Biosym suite of software (Insight II), O, and manual docking algorithms
were employed to identify tricyclic ring systems that have the
appropriate geometry and van der Waals complimentarity.39 Candidates
including hetero-tricyclic ring systems, especially fused 6-6-6 and
6-5-6 systems were further scrutinized on the basis of which had the
appropriate geometry and functionality coupled with applicable synthetic
methodology to direct a phenyl substructure into the inner binding site.
Another design feature considered was ease of synthesis for the
introduction of the carboxylate substituents, such that electrostatic
interactions with the Lys-15 ammonium groups in TTR could contribute
to binding affinity. A few ring systems were identified, of which the
phenoxazine class appeared ideal.
Sedimentation Velocity Analysis. All sedimentation analyses were
performed on samples used previously for turbidity analysis, which
measures the extent of amyloid fibril formation after 72 h or 7 days in
100 mM acetate buffer, 100 mM KCl, 1mM EDTA (pH 4.4).
Analysis of Sedimentation Velocity Profiles of TTR. The sedi-
mentation properties of WT and L55P TTR solutions, in the presence
and absence of inhibitors, were analyzed on a temperature controlled
Beckman XL-I analytical ultracentrifuge equipped with a An60Ti rotor
and a photoelectric scanner. Data were collected at speeds of 3000 and
50 000 rpm in continuous mode at 20 °C, employing a step size of
0.001 cm. Detection was carried out at 280 nm for samples incubated
for 6 days and at 230 nm for the samples incubated for 3 days.
A direct boundary fitting approach was applied to evaluate the
sedimentation velocity data derived from a 3.6 µM solution of TTR
incubated with a 7.2 µM solution of 4 at pH 4.4. The Svedberg program
was used to fit multiple concentration vs radial position data sets
simultaneously to yield approximate solutions to the Lamm equation.44
The fitting algorithm yields the sedimentation coefficient and diffusion
coefficient which affords the molecular weight using equation I.
1
(57%); H NMR (400 MHz, DMSO-d6) δ 6.65 (d, J ) 7.6 Hz, 2H),
6.84 (t*, J ) 7.6, 7.9 Hz, 2H), 7.03 (d, J ) 7.9 Hz, 2H), 8.70 (s, 1H);
13C NMR (100 MHz, DMSO-d6) δ, 116.75, 118.74, 122.00, 124.29,
132.71, 141.53, 165.48; HRMS (DHB) m/z 294.0377 (M + Na)+
requires C14H9NO5Na requires 294.0378).
10-(m-Trifluoromethylpyrimidine)-phenoxazine-4,6-dicarboxy-
late (20). Dimethyl ester 34 (17.1 mg, 0.05 mmol) was subjected to
the ester hydrolysis procedure outlined above. The crude product was
dissolved in DMSO and purified by reverse phase HPLC, employing
a linear gradient of 50-100% B over (retention time ) 15 min) 25
1
min to yield 15 mg of 20 (98%); H NMR (400 MHz, DMSO-d6) δ
7.32 (t, J ) 7.9 Hz, 2H), 7.57 (d, J ) 5.0 Hz, 1H), 7.62 (dd, J ) 7.9,
1.5 Hz, 2H), 7.98 (dd, J ) 7.9, 1.5 Hz, 2H), 8.86 (d, J ) 5.0 Hz, 1H)
13C NMR (100 MHz, DMSO-d6) δ 111.01, 118.94, 121.95, 123.20,
127.68, 129.95, 149.35, 154.67, 158.77, 161.98, 165.52; MALDIFTMS
(DHB) m/z 440.0466 (M + Na)+ requires C21H15NO6Na requires
440.0466).
General Procedure for Phenol Triflation. The triflation procedure
previously described by Stille et al., was used to synthesize all triflates
from their respective phenols.43 Briefly, a 10 mL round-bottom flask,
fitted with a septum and a dry stir bar, was charged with the phenol (9
mmol), and distilled pyridine (3 mL, Aldrich, the solvent) via syringe
through the septum. The reaction mixture was cooled to 0 °C with an
ice/H2O bath. To initiate the reaction trifluoromethane sulfonate
anhydride (12 mmol, Aldrich) was then added via syringe through the
septum. The ice bath was removed, and the reaction was allowed to
warm to room temperature and stir overnight. The reaction mixture
was then poured into a 250 mL beaker containing 50 mL of an ice/
H2O slurry. The contents of the beaker were transferred to a 125 mL
separatory funnel, and the aqueous layer was extracted with Et2O (3 ×
35 mL). The organic layers were combined and dried over MgSO4,
and the solvent was removed under reduced pressure. The triflates were
purified via flash chromatography from the resultant oil.
3-Methylphenyl Trifluoromethylsulfonate (35). m-Cresol (1 g, 1.54
mL, 9.2 mmol) was subjected to the general triflation procedure outlined
above. The crude product was purified by chromatography (300 mg of
SiO2, 20% EtOAc in hexanes, Rf ) 0.75) to afford 2.1 g (93%) of 35
as a pale yellow oil; 1H NMR (600 MHz, CDCl3) δ 2.34 (s, 3H), 7.02-
7.07 (m, 2H), 7.15 (m, 1H), 7.27 (m, 1H); 13C NMR (150 MHz,CDCl3)
δ 20.96, 118.12, 121.71, 129.11, 129.83, 140.92, 149.61; GCMS
(electron ionization) m/z 240 (C8H7F3O3S requires 240).
3-Ethylphenyl Trifluoromethylsulfonate (36). 3-Ethylphenol (1 g,
1.001 mL, 8.2 mmol) was subjected to the general triflation procedure
outlined above. The crude product was purified by chromatography
(300 mg of SiO2, 20% EtOAc in hexanes, Rf ) 0.79) to afford 1.81 g
(87%) of 36 as a pale yellow oil; 1H NMR (600 MHz, CDCl3) δ 1.22
(t, J ) 7.9 Hz, 3H), 2.67 (q, J ) 7.9 Hz, 2H), 7.05-7.10 (m, 2H),
7.24 (m, 1H), 7.31 (m, 1H); 13C NMR (150 MHz,CDCl3) δ 15.00, 28.51,
118.36, 120.58, 127.92, 129.95, 147.24, 149.73; GCMS (electron
ionization) m/z 254 (C9H9F3O3S requires 254).
3-Isopropylphenyl Trifluoromethylsulfonate (37). 3-Isopropyl-
phenol (0.994 g, 1.00 mL, 7.2 mmol) was subjected to the general
triflation procedure outlined above. The crude product was purified by
chromatography (300 mg of SiO2, 5% EtOAc in hexanes, Rf ) 0.43)
to afford 1.33 g (69%) of 37 as a pale yellow oil; 1H NMR (600 MHz,
CDCl3) δ 1.23 (d, J ) 7.0 Hz, 6H), 2.91 (heptet, J ) 7.0 Hz, 1H),
7.06 (d, J ) 7.9 Hz, 1H), 7.11 (s, 1H), 7.22 (d, J ) 7.9 Hz, 1H), 7.30
(t, J ) 7.9 Hz, 1H); 13C NMR (150 MHz,CDCl3) δ 23.45, 33.94, 118.47,
119.24, 127.12, 129.98, 149.83, 151.98; GCMS (electron ionization)
m/z 268 (C10H11F3O3S requires 268).
MW ) sRT/D(1-υjF)
(I)
In equation I, MW is the molecular weight in (Da), s is the
sedimentation coefficient (in Svedberg units, 10-13 sec), R is the
universal gas constant (8.314 × 107 erg/mol), υj is the partial specific
volume (cm3 / g), and F is the solvent density (g/cm3). The buffer density
(1.00848 g/cm3) was calculated from tabulated data. The partial specific
volumes of WT (0.7346 cm3/g) and L55P TTR (0.7334 cm3/g) were
calculated from the amino acid composition.11
Analysis of Sedimentation Equilibrium Properties of TTR under
Fibril Formation Conditions in the Presence of Phenoxazine-Based
Inhibitors. Sedimentation equilibrium measurements were made using
120-140 µL of TTR solution (3.6 µM) loaded into a double sector
cell, equipped with a 12-mm Epon centerpiece and sapphire or quartz
windows. Data were collected initially at rotor speeds of 3000 rpm to
ensure the absence of large molecular weight oligomers (these would
sediment) and then at 17 000 rpm to allow equilibrium to be established
across the cell. The sedimentation profiles monitored (280 nm) at 3 h
intervals were overlaid to establish that equilibrium had been achieved.
Data analysis was carried out using a nonlinear least-squares analysis
in the Origin software package provided by Beckman. Several different
models including a single ideal species model and several multiple
species models were fit to the data to identify the simplest model that
best fits the data. Equation II corresponding to a single ideal species
model fits the data best based on the small differences between the
theoretical data and the experimental data:
2
Ar ) exp[ ln(Ao) + (Mω2 (1-υjF/RT)(x2 - xo )] + E
(II)
where Ar is the absorbance at radius x, Ao is the absorbance at a reference
radius xo (usually the meniscus), ω is the partial specific volume of
the protein, F is the density of the solvent (g/cm3), ω is the angular
velocity of the rotor (radian/sec), E is the baseline error correction factor,
M is the molecular weight, and R is the universal gas constant. The
differences between the experimental data points and the fitted data
Structure-Based Ligand Design. The structure of the TTR‚(Flu)2
complex served as a starting point for the design of inhibitory ligands.
Visual inspection of the electron density traced out by both symmetry
equivalent binding modes of Flu in the outer hormone binding cavity
(43) Echavarren, A. M.; Stille, J. K. J. Am. Chem. Soc. 1987, 109, 5478-
5486.
(44) Philo, S. J. In Modern Analytical Ultracentrifugation; Shuster, T.
M., Laue, T. M., Eds.; Birkha¨user: Boston, 1994; pp 156-170.

2186 J. Am. Chem. Soc., Vol. 122, No. 10, 2000
Petrassi et al.
Table 3. Crystallographic Analysis of TTR‚(4)2 Complexa
points (the residuals) are randomly distributed and small in magnitude,
other models did not fit the data well as discerned by the nonrandom
distribution of residuals across the cell.
resolution, Å
reflections, measured/unique
completeness, % overall/outer shell
10-1.8
75643/22052
95.3/97.5
4.4/17.3
18.0
Fibril Formation Assay. Recombinant WT and L55P TTR were
purified from Escherichia coli as described previously.25 The stagnant
amyloid fibril formation assay recently published by our laboratory
was used to assess the efficacy of the phenoxazine-based amyloid fibril
inhibitors.10,32,35,36 Briefly, stock solutions of each inhibitor were
prepared by dissolving the potential inhibitor in DMSO to a concentra-
tion of 7.2 mM (inhibitors were dried in a vacuum desiccator (<0.01
mmHg) for 12 h in the presence of phosphorus pentoxide and the
solutions were made by weight). For assays in which the final inhibitor
concentration was 7.2 or 3.6 µM, the initial 7.2 mM inhibitor stock
solution was diluted to 1.4 mM and 720 µM solutions respectively
with DMSO. A 7.2 µM TTR stock solution was prepared by dilution
from a 3-5 mg/mL stock solution that had been dialyzed against a 10
mM phosphate buffer, 100 mM KCl, 1 mM EDTA (pH 7.6). The
protein solution (495 µL) was delivered to Eppendorf tubes for each
triplicate measurement, to which 5 µL of the inhibitor solution was
added (except for the controls). Promising phenoxazines were tested
at least three times in triplicate.
R
sym, % overall/outer shell
I/σI
Refinement Statistics
resolution, Å
completeness, %
R
8-1.9
82.4 (F > 3σ)
20.2/20.3
0.008
factor/Rfree, %
rmsd bond length, Å
rmsd bond angle, deg
1.2327
a Rsym ) Σ|I - I |/Σ I , where I is the observed intensity, and I is
the average intensity of multiple observations of symmetry-related
reflections. Rfactor ) Σ|Fo- Fc|/ΣFo where Fo and Fc are observed and
calculated structure factor amplitudes, respectively.
CIFTAB to generate the tables. Final atomic coordinates and selected
bond lengths and angles are tabulated in the Supporting Information
Tables.
X-ray Crystallographic Determination of the TTR‚(4)2 Complex.
Crystals of recombinant TTR were grown from solutions of 5 mg/mL
TTR (in 100 mM KCl, 100 mM phosphate, pH 7.4, 1 M ammonium
sulfate) equilibrated against 2 M ammonium sulfate in hanging drop
experiments. The complex of TTR with 4 was prepared from crystals
soaked for 5 weeks with a 10-fold molar excess of the ligand to ensure
full saturation of both binding sites. A DIP2030 imaging plate system
(MAC Science, Yokohama, Japan) coupled to an RU200 rotating anode
X-ray generator was used for data collection. Crystals of the TTR‚(4)2
complex are isomorphous with the native crystal form with unit cell
dimensions in the range of a ) 43 Å, b ) 86 Å, and c ) 65 Å. They
belong to the space group P21212, and contain one-half of the homo-
tetramer in the asymmetric unit. Data were reduced with DENZO and
SCALEPACK.
The inhibitor-TTR solutions were preincubated for 30 min at 37
°C to allow plenty of time for potential inhibitors to bind to TTR. The
TTR solutions were rendered acidic and amyloidogenic by the addition
of 500 µL of 200 mM acetate buffer, 100 mM KCl, 1 mM EDTA (pH
4.2), which was added to each TTR solution to yield a final pH of 4.4.
These solutions were incubated for 72 h (37 °C) or a 7 day incubation
(pH 4.4) to evaluate inhibitor efficacy. The Eppendorfs were vortexed
for 5 s and their optical density at 400 nm measured to quantify fibril
formation. The extent of TTR fibril formation in the absence of inhibitor
was defined to be 100% (typically an o.d. between 1.15 and 1.25 for
WT (60% chemical yield of fibrils), hence active inhibitors give <100%
yield of fibrils. Inhibitors were tested in the absence of TTR to evaluate
their intrinsic absorbance and to be sure that the inhibitor is soluble
over the course of the assay (i.e. does not contribute to the turbidity,
none of the phenoxazines show a o.d. > 0.05). Flufenamic acid (7.2
µM), (Table 2) was included as a positive control which typically yields
4 ( 5% of WT TTR fibril formation (95% inhibition).33
The protein atomic coordinates for apo TTR from the Protein Data
Bank (accession number 1BMZ) were used as a starting model for the
TTR‚(4)2 complex by molecular dynamics and energy minimization
using the program CNS. Simulated annealing and individual temperature
factor refinement provided a model that was used to phase the
“complex-native” difference Fourier maps. In the resulting maps, the
ligands could be located in both binding pockets of the TTR tetramer
with peaks heights of above 7 σrms. At this point the ligand molecule
was fit into the density found in both hormone binding cavities and
was included into the crystallographic refinement. The conformation
of the bis-methyl ester of 4, 22, determined by small molecule X-ray
diffraction (Figure 1D, and Supporting Information) was in good
agreement with the initial annealed 2 |Fo,complex| - |Fo,apo| Φapo maps
and was used as initial conformation of 4 in the crystallographic
refinement. Because of the 2-fold crystallographic axis along the binding
channel, a statistical disorder model had to be applied, giving rise to
two ligand binding modes in each of the two binding sites of tetrameric
TTR. After several cycles of simulated annealing, subsequent positional
and temperature factor refinement placed 60 H2O molecules into the
difference Fourier maps. For the TTR‚(4)2 complex both molecular
conformations of 4 were in good agreement with unbiased annealed 2
|Fo| - |Fc| Φcalc omit maps, phased in absence of the inhibitor. Because
of the lack of interpretable electron densities in the final map, the 9
N-terminal and 3 C-terminal residues were not included in the final
model. A summary of the crystallographic analysis can be found in
Table 3. Coordinates for the TTR‚(4)2 complex have been deposited
in the Brookhaven Protein Data Bank (accession numbers 1DVY).
Isothermal Titration Calorimetry. A filtered solution of 4 in 10
mM phosphate buffer, 100 mM KCl, 1 mM EDTA (pH 7.6) was titrated
(25 °C) into a 1.3407 mL solution of TTR in the same buffer using a
Microcal MCS isothermal titration calorimeter (Microcal Inc., Northamp-
ton, MD). Compound 4 was manipulated by dissolving reverse phase
HPLC purified material into 5% ammonium carbonate solution which
was lyophilized to afford the ammonium salt of 4. The solution of 4
was then filtered through a 0.2 µ cellulose filter and its concentration
determined by absorbance at 323 nM (ꢀ ) 6726 M-1 cm-1). The TTR
solution was prepared by dialysis of ∼4 mg/mL solution against the
Structural Determination of 22 via X-ray Crystallography. A
single parallelpiped shaped crystal of compound 22 was obtained by
slow evaporation of a concentrated solution in CHCl3/hexanes. The
crystal was mounted along the largest dimension and data collected
with a Rigaku AFC6R diffractometer equipped with a copper rotating
anode and a highly orientated graphite monochromator. A constant scan
speed of 8 °/min in a ω was used and weak reflections [I > 5σ > I]
were rescanned a maximum of 6 times and the counts accumulated to
ensure good counting statistics. The intensities of three monitored
reflections measured after every 200 reflections did not change
significantly during the 38 h of X-ray exposure. Unit cell dimensions
and standard deviations were obtained by least-squares fit to 25
reflections (80° < 2Θ < 100°). The data were corrected for Lorentz
and polarization effects, but not for absorption because of low value
of µ.
There were no systematic absences in the data, and therefore the
space group P1 was assumed and later confirmed by successful
refinement of the structure. The structure was solved by direct methods
using SHELX-97. All non-hydrogen atoms were refined anisotropically
by the full matrix least-squares method. The function minimized was
2
Σw (||Fo|| - ||Fc|| ). Hydrogen atoms were included in the ideal
positions with fixed isotropic U values equal to 1.2 times that of
the atom they are attached to. A weighting scheme of the form w )
1/[σ(Fo ) + (aP)2 + bP] with a ) 0.0945 and b ) 0.598 was used. P
2
2
is defined as Max ((Fo , 0) + 2 Fc2)/3). An extinction correction was
also applied to the data. The refinement converged to the R indices
given in supplemental Table 1. The final difference map was devoid
of significant features.
All calculations were done on a Silicon graphics personal Iris 4D35
and an IBM-compatible PC using the program TEXSAN (data
reduction), SHELX-97 (refinement), SHELXTL-PC (plotting) and

Design of TTR Amyloid Fibril Inhibitors
J. Am. Chem. Soc., Vol. 122, No. 10, 2000 2187
the TTR‚(T4)2 structure.33,39 Meta substituents on the inner
phenyl ring should also mediate the Ser-117 side chain
rearrangement described in the Introduction, which may stabilize
the TTR tetramer, owing to intersubunit H-bonding.33 Therefore
m-substituted N-phenyl phenoxazines and related analogues were
used to interrogate the inner TTR binding cavity in an effort to
optimize the structure required to best stabilize the TTR tetramer
against dissociation and tertiary structural changes required for
amyloidogenesis.
Synthesis of Phenoxazines. The synthesis of the N10-phenyl-
4,6-phenoxazine dicarboxylic acids desired for this study
commences with protection of N10 in commercially available
phenoxazine, Figure 2. Following the procedure of Antonio et
al., the anilino proton was removed by NaH and the N-centered
anion silylated with tert-butyldimethyl silyl(TBDMS) chloride
affording 1.40 The sterically demanding TBDMS group prevents
N-directed metalation, allowing for selective oxygen directed
ortho metalation at the 4,6-positions utilizing n-BuLi. Bis-
lithiated phenoxazine was treated with gaseous CO2, followed
by esterification of the resulting dicarboxylic acid with trim-
ethylsilyldiazomethane to yield the 4,6-dimethyl ester 2. Re-
moval of the N-silyl-protecting group with tetrabutylammonium
fluoride generates anilino diester 3. The anilino diester (N10)
was then coupled to a variety of aromatic iodides, bromides, or
triflates using the Pd-mediated N-aryl coupling reactions
developed by Buchwald and Hartwig, Figure 2 (see Table 1
for a summary of the analogues prepared). 41,42 In cases where
the aryl triflate was not commercially available, it was prepared
by treating the respective phenol with triflic anhydride in
pyridine.43 The methyl esters in compounds 21-29 and 31-34
were hydrolyzed using LiOH. In the case of 30, where a methyl
ether and methyl esters comprise the same molecule, the methyl-
protecting groups were removed simultaneously by BBr3
treatment.
Figure 2. Summary of the key synthetic transformations in the
synthesis of the N10-phenyl-4,6-phenoxazine dicarboxylic acids; (a)-
NaH, TBDMSCl, (b) n-BuLi, (c) gaseous CO2, (d) TMSCHN2, (e)
TBAF, (f) Pd2(dba)3, Cs2CO3, (()Binap, (g) LiOH or BBr3.
same phosphate buffer, filtering as above. The final concentration was
determined by absorbance at 280 nm. In the analysis of the binding of
4 to WT TTR (13.8 µM), the initial injection of 2 µL (530 µM) was
followed by 24 injections of 8 µL, allowing a 4 min equilibration period
between each injection. In the case of L55P TTR (23 µM) the initial
injection of 2 µL of 4 (570 µM) was followed by 24 injections of 10
µL, with a 6 min equilibration period between injections. Three
independent titrations for both WT and L55P TTR were performed.
Two sets of blanks were also collected in triplicate, the first set of
blanks being the titration of each solution of 4 into buffer (heat of
dilution of the ligand) and the second being the titration of buffer into
the respective protein solutions (heat of dilution of the protein).
Integration of the thermogram and subtraction of all blanks gives a
binding isotherm that fits to a model of two interacting sites exhibiting
negative cooperativity. The data were fit by using the variables, K1,
H1, K2, and H2 and the ITC Data Analysis Module in ORIGIN version
2.9 provided by Microcal.
Results
Amyloid Fibril Inhibition. The extent of transthyretin
amyloid fibril formation (from a 3.6 µM solution of either WT
or L55P TTR at pH 4.4) in the presence of phenoxazine-based
inhibitors 4-20 (3.6 or 7.2 µM) is reported in Table 1. The
activities of compounds 4-6, 17, and for comparison purposes,
flufenamic acid are represented in bar graph format in Figure 3
A and B. Compounds 4 and 5 were chosen because they are
the best inhibitors, Flu is shown for comparative purposes as it
is the lead molecule upon which the phenoxazine design was
based. Compounds 6 and 17 were chosen for bar graph
representation because they are similar in structure to 4 and 5,
but have surprisingly poor activity. When comparing inhibitors,
percent conversion of tetrameric TTR into amyloid and its
precursors (oligomers, protofilaments, filaments) over a 72 h
period is reported relative to TTR lacking inhibitor (100%).
Hence, 100% fibril formation reflects no inhibition, whereas
0% fibril formation is associated with complete amyloid fibril
inhibition.
Design of Phenoxazines. The anthranilic acid substructure
of Flu occupies only about 50% of the outer TTR thyroid
hormone binding cavity in each symmetry equivalent binding
mode, Figure 1, A and B. Consideration of the composite
electron density (Figure 1B) occupied by both binding modes
of Flu, (transformed by the crystallographic C2z axis) suggests
that a tricyclic ring system such as phenoxazine (Figure 1C)
should fill the entire outer hormone binding cavity.33 The
geometry of phenoxazine appears to perfectly complement the
van der Waals surface of the outer TTR binding cavity, which
should lead to an increase in binding affinity and selectivity
relative to Flu.39 Furthermore, the N10 of phenoxazine (Figure
2) is ideally positioned to direct a substituted phenyl ring into
the inner hormone binding cavity, Figure 1C. Hence, an N10-
phenyl phenoxazine appears optimal to fill both the outer and
inner TTR binding cavities simultaneously, Figure 1C. All of
the active inhibitors discovered by screening have carboxylate
or phenolic substituents on the aromatic ring occupying the outer
TTR binding cavity.35-37,39,45 From unpublished crystallographic
results and from the published TTR‚(Flu)2 structure outlined in
the Introduction, we know that these acidic groups are ideally
positioned to interact with the ꢀ-ammonium groups of Lys-15,
projecting into the periphery of the outer thyroid binding cavity
from two adjacent TTR subunits.33 Therefore, N-phenyl phe-
noxazines substituted with carboxylate groups at the 4- and
6-positions (Figure 2) should bind with high affinity due in part
to these electrostatic interactions, Figure 1C. The m-CF3
substituted phenyl ring of Flu nearly completely fills the van
der Waals surface of the inner binding cavity of TTR, including
one of the halogen binding pockets occupied by an iodine in
Phenoxazines 4 and 5 are better than Flu at inhibiting L55P
TTR amyloid fibril formation, at a small molecule concentration
of either 3.6 (Figure 3A) or 7.2 µM, Figure 3B. Compounds 4
and 5 are similar to Flu with respect to the inhibition of WT
amyloid formation, Figure 3, A and B. Phenoxazines 6-10 are
equal or better than Flu against the formation of L55P amyloid
fibrils; however, 3.6 µM solutions of 6-10 are slightly inferior
to Flu from the standpoint of inhibiting WT amyloidogenesis,
Table 1. It is generally the case that meta-substitued N-
phenylphenoxazines are superior to para-substitued N-phenyl
phenoxazines, which are better than the unsubstituted N-phenyl
phenoxazine, 17, Table 1. The 4,6-phenoxazine dicarboxylate,

2188 J. Am. Chem. Soc., Vol. 122, No. 10, 2000
Petrassi et al.
effects of ligand binding to a protein with two identical sites
(non-cooperative binding). In contrast to the similarities ob-
served in the binding constants of 4 and Flu to WT TTR, 4
binds to the two sites of L55P TTR with decreased affinity in
comparison to Flu. Phenoxazine 4 binds to the fist site of L55P
with a Kd1 of 357 nM. This is over 5 times the Kd1 characterizing
Flu binding to L55P. The differences in the dissociation constant
of 4 and Flu binding to L55P TTR at the second site are not as
dramatic (1.5-fold), Table 2A. The negative cooperativity
associated with L55P ligand binding disappears when comparing
Flu to 4, very similar to the trend observed for WT TTR.
Dissection of the free energies associated with phenoxazine
4 and Flu binding to TTR reveals a noteworthy difference.
Flufenamic acid binding to both sites in WT TTR is enthalpi-
cally driven, with the entropy associated with the binding and
linked TTR conformational changes being unfavorable, Table
2B. On the contrary, there is a favorable entropic component
for the binding and linked conformational changes associated
with the interaction of 4 with WT TTR, Table 2B. The entropy
term contributes 53% (5.15 of 9.67 kcal/mol) of the free energy
for the first binding event and 43% (3.85 of 9.03 kcal/mol) of
the free energy for the second binding event, the remaining
contribution coming from enthalpy. A favorable entropy con-
tribution to the free energy is also observed in the case of the
binding of 4 to L55P TTR, (Table 2C) although in this case the
enthalpy terms make the dominant contribution to the binding
free energy.
Figure 3. A bar graph representation of the relative yield of TTR
amyloid precursors and fibrils produced from WT and L55P TTR (3.6
µM) at pH 4.4, after 72 h of incubation at 37 °C in the presence of
either 3.6 (A) or 7.2 µM (B) concentration of inhibitor. When comparing
inhibitors percent conversion of tetrameric TTR into amyloid is reported.
Hence, 100% fibril yield reflects no inhibition, whereas a 0% fibril
yield corresponds to complete inhibition.
Analytical Equilibrium Ultracentrifugation Studies. Sedi-
mentation equilibrium and sedimentation velocity experiments
evaluating the quaternary structure of WT and L55P TTR (3.6
µM) were performed with and without inhibitor 4 (7.2 µM), at
a pH associated with maximal amyloid fibril formation (pH 4.4).
WT TTR bound to 4 remains tetrameric after 72 h of incubation
(pH 4.4) with no detectable loss of protein (<5%) according to
sedimentation equilibrium studies, Figure 4A. The fit to a single
ideal species model is excellent as evident from the small
magnitude and random distribution of the residuals. The solution
molecular weight from the single ideal species model is 51 513
( 430 Da, in good agreement with the expected MW of the
TTR tetramer (55 032 Da). WT TTR self-assembles into high
MW oligomers that sediment from the cell in the absence of
inhibitor (72 h of incubation, pH 4.4, 37 °C) at the same angular
velocity (17 000 rpm) used for equilibrium experiments de-
scribed directly above (data not shown). Similarly, in the absence
of inhibitor 4, L55P TTR (3.6 µM) forms large oligomers that
sediment out of the cell (17 000 rpm) under amyloid fibril
forming conditions (pH 4.4, 37 °C, 72 h) (data not shown).
However, in the presence of inhibitor 4, an equilibrium analysis
(17 000 rpm) reveals retention of >95% of the tetrameric form
of L55P TTR, even after 7 days of preincubation at pH 4.4,
Figure 4B. The solution molecular weight from the ideal species
model is 50 110 ( 476 Da, in very good agreement with the
expected MW of the tetramer (54 982 Da). Sedimentation
velocity studies of WT and L55P TTR (3.6 µM) preincubated
for 6 days at pH 4.4 in the presence of a 7.2 µM solution of 4
also demonstrate that TTR remains tetrameric. In both cases,
TTR sediments as a single species with an S valve of 3.7 ( 0.1
(Supporting Information, Figures 3 and 4) corresponding to
tetrameric TTR. It is clear that N-phenyl phenoxazine 4 stabilizes
the TTR tetramer under acidic conditions where it would
otherwise efficiently dissociate to monomers and self-associate
into oligomers, ultimately forming amyloid filaments and
fibrils.10,32,33 The off rate of the inhibitor must be very slow
relative to the on rate and the rate of tetramer dissociation, as
19, is not by itself capable of significant amyloid fibril inhibition,
demonstrating the importance of N-phenyl ring. An analogue
of 6 lacking both carboxylate groups is also inactive as an
inhibitor, even at a concentration of 36 µM (data not shown),
demonstrating the importance of the electrostatic interactions
for binding and inhibition. However, we know from 19 that
these interactions alone are not sufficient for inhibition.
Compound 20 is unique with respect to the other N-phenyl
phenoxazines outlined in Table 1 in that the pyrimidine ring
can rotate freely, relative to the phenoxazine ring system, Table
1. The conformational freedom exhibited by phenoxazine 20
did not improve biological activity; in fact, it is a very poor
inhibitor, Table 1.
Binding Constant Determination. Isothermal titration cal-
orimetry was employed to measure the binding constants of the
phenoxazine-based TTR amyloid fibril inhibitor 4 and to
evaluate the cooperativity or lack thereof between the TTR
binding sites, Table 2. The raw isothermal titration calorimetry
data and the fit to the isotherms for 4 binding to WT and L55P
TTR can be found in Supporting Information, Figures 1 and 2,
respectively. Phenoxazine 4 exhibits binding to WT TTR that
is similar to that of flufenamic acid. The dissociation constant
(Kd1) of 4 for the first site of WT TTR is 79 nM, (Table 2A) a
2-fold decrease in affinity relative to Flufenamic acid (Kd1
)
30 nM). The binding affinities of 4 and Flu for the second site
of WT TTR are within error (237 ( 26 vs 255 ( 97 nM), Table
2A. Hence, the negative cooperativity exhibited for Flu binding
is not observed for 4 binding to TTR, as the ≈4-fold difference
observed between Kd1 and Kd2 is that expected for the statistical

Design of TTR Amyloid Fibril Inhibitors
J. Am. Chem. Soc., Vol. 122, No. 10, 2000 2189
ring system of 4 spans the entire outer hormone binding cavity
of TTR, making van der Waals interactions with the side chains
of Thr-106, Lys-15, and Leu-17 in two adjacent TTR subunits
simultaneously, Figure 5, A and B. This binding conformation
positions one of the carboxylate substituents of 4 in close enough
proximity to form a weak hydrogen bond with the ꢀ-ammonium
group of Lys-15 (O4:Nꢀ distance 3.3 Å), Figure 5B. The second
carboyxlate group is 3.2 Å from the carboxylate of Glu-54
allowing for a possible weak hydrogen bond, assuming that one
of the carboxylate groups is protonated at pH 7.4, Figure 5B.
The binding of the m-CF3-substituted phenyl ring of 4
displaces the ordered H2O molecule found in the innermost
binding cavity of the apo TTR structure. Hydrophobic interac-
tions are observed between m-CF3 group of 4 and the second
most interior halogen binding pocket, delineated by the side
chains of Leu-17, Leu-110, and Thr-119, Figure 5B, Supporting
Information Figure 8. Binding of 4 causes the hydroxyl side
chains of all four Ser-117 residues to rotate 120° toward the
interior of the TTR tetramer. The rotated Ser-117 side chains
appear to form the intersubunit H-bonding interactions observed
in the TTR‚(Flu)2 structure, but not observed in the structure
of apo TTR, Figure 5B.33 The m-CF3 phenyl substructure of 4
does not bind as deeply into the inner binding cavity as the
corresponding group in the TTR‚(Flu)2 structure (carbon atom
of CF3 group is translated outward by 2.5 Å), compare
Supporting Information Figures 7 and 8.33
Discussion
The focus of this study was to use a structure-based ligand
design strategy to produce small molecule inhibitors against WT
and L55P transthyretin amyloid fibril formation. The extreme
amyloidogenicity of L55P TTR and the poor ligand binding
characteristics generally exhibited by this variant represent a
demanding target for scrutinizing potential small molecule
inhibitors (see introduction). The design criteria outlined in the
results section were tested by synthesizing 15 N-phenyl-4,6-
phenoxazine dicarboxylates.
The experimental results summarized in Table 1 and Figure
3, A and B, demonstrate that several N-phenyl phenoxazines
are very good inhibitors against both WT and L55P amyloid
fibril formation in Vitro. Phenoxazines 4 and 5 are excellent at
inhibiting both WT and L55P TTR amyloid fibril formation.
The ability of 4 and 5 to completely inhibit L55P fibril formation
at a concentration of 7.2 µM is unprecedented, Figure 3B. The
analogue studies outlined in the results section demonstrate that
the entire structure of 4 and 5 appear to be required for inhibitor
efficacy. That the substituted N-phenyl ring is essential can be
discerned from the poor activity of phenoxazines 19 and 17.
Inhibitor 6 demonstrates that bis-meta substitution does not
improve activity; instead, activity decreases. An analogue of 6
lacking both carboxylates is also a very poor inhibitor, even at
a concentration of 36 µM, demonstrating the importance of the
electrostatic interactions. The numerous active meta- and para-
substituted N-phenyl-4,6-phenoxazine dicarboxylates (e.g., 4-10,
Table 1) suggest that additional manipulation of this N-phenyl
ring could yield inhibitors that are even more potent.
Figure 4. Sedimentation equilibrium analysis of 3.6 µM solutions of
WT TTR (A) and L55P TTR (B) in the presence of 4 (7.2 µM) at pH
4.4 (maximal fibril formation conditions) after 72 h and 7 days of
incubation, respectively. The solid line was derived from fitting the
absorbance vs the radial position data to Equation II for a homogeneous
tetrameric species bound to 2 equiv of 4. The residual difference
(residuals) between the experimental data (O) and the fit at each data
point is shown in the inset.
fibril formation does not noticeably increase when a 72 h time
point is compared to a 168 h time point.
Solid State Characterization of Inhibitor 4. The bis-methyl
ester of phenoxazine 4 (compound 22) was crystallized to
characterize its preferred conformation. Not surprisingly, the
N-phenyl ring is oriented roughly perpendicular (86°) to the
plane of the phenoxazine ring system, Figure 1D, Supporting
Information Figures 5 and 6. The phenoxazine ring system itself
is nearly flat, exhibiting only a 11.3° angle between the two
phenyl rings composing the tricyclic ring system. The N10 bond
angles (118.2°, 118.9°, 119.6°) and planar orientation of the
attached ligands are consistent with sp2 N10 hybridization
(Supporting Information Tables). Knowing the preferred con-
formation allows for a more thorough evaluation of the TTR‚
(4)2 cocrystal structure discussed below.
The TTR‚(4)2 X-ray crystal Structure. The 1.9 Å resolution
structure of 4 bound to both binding sites within TTR
demonstrates that the N-phenyl phenoxazine binds in its low-
energy conformation in two symmetry equivalent modes, Figure
5A. Both were in good agreement with the unbiased 2 |Fo| -
|Fc| Φcalc omit maps phased in the absence of 4. The tricyclic
Phenoxazine 4 binds to WT TTR with a dissociation constant
of 79 ηM, 2-fold higher than that exhibited by Flu, whereas
the dissociation constant for Flu and 4 (237 ηM) binding to the
second TTR site are within error, Table 2A. Hence, it is not
surprising that Flu and 4 exhibit very similar efficacy as WT
TTR amyloid fibril inhibitors, since the concentrations employed
(3.6 or 7.2 µM) are significantly higher than the dissociation
constants, owing to the concentration of TTR in human plasma

2190 J. Am. Chem. Soc., Vol. 122, No. 10, 2000
Petrassi et al.
Figure 5. Structural representations of the X-ray structure of 4 bound to one of the two C2 symmetrical TTR hormone binding sites. (A) Framework
model representation of the two symmetry equivalent binding modes of 4 (one where C atoms are depicted in yellow and the other in aqua). Only
residues from one subunit of TTR that form the binding cavity are displayed with their 2 |Fo| - |Fc| Φcalc electron density in cyan. The symmetry
related binding modes of 4 fit nicely into the final 2 |Fo| - |Fc| Φcalc electron density in green. In the framework model, oxygen atoms are in red,
nitrogen atoms in blue, and fluorine in green. The protein ligand interactions are described in the text. (B) A molecular surface representing a cut
away view of one of the TTR binding pockets (shown in gray) occupied by 4 in one of its symmetry equivalent binding modes. Ligand 4 is depicted
in stick format using the same color scheme as in A, with the exception that carbons are in black. The binding site is formed by two adjacent TTR
monomers related by a C2z operation (see Figure 1A). Residue side chains undergoing rearrangement upon binding 4 include Ser-117, Thr-119, and
Lys-15. The m-CF3 substituted phenyl ring of Flu is translated 2.5 Å deeper into the innner binding sub-site that the analogous substructure in 4
(see Supporting Information Figures 7 and 8 for a comparison). Interactions of 4 with the periphery of the TTR binding site likely maximize
hydrophobic interactions and allow some interactions of 4 with D strand residues such as Glu-54, which may prevent the putative C-strand-loop-
D-strand rearrangement that leads to the formation of the amyloidogenic intermediate.
(3.6 µM). The binding of 4 to L55P TTR is characterized by a
5-fold increase in Kd1 (357 ηM) and 1.5-fold increase in Kd2
(1050 ηM) relative to Flu, Table 2A. Yet, phenoxazine 4 is a
better inhibitor than Flu when used at a concentration of either
3.6 or 7.2 µM. This apparent conundrum can be rationalized
by considering the possibility that inhibitor efficacy results from
both ground-state and transition-state effects. The possibility
that 4 raises the free energy of the transition state associated
with the tetramer to amyloidogenic monomer transition in
addition to stabilizing the L55P TTR‚(4)2 ground state is not
unreasonable based on the Hammond Postulate, which predicts
that the transition state structure would be tetramer-like. Hence
ligand binding could destabilize the transition state, Figure 6.
In addition, it is possible that the binding constants of Flu and
4 could reverse their relative affinities over the pH range of
7.6-4.4, providing an alternative explanation for the lack of
correlation between binding and inhibitor efficacy, the latter
limit being the pH where the inhibitor evaluations were made
(37 °C) and the former where the binding constants were
measured by ITC experiments (25 °C). Unfortunately, dissocia-
tion and conformational changes leading to amyloid fibril
formation preclude analyzing binding at pHs lower than 7.6 (a
detailed summary of the technical problems can be found
elsewhere).32
The 1.9 Å X-ray crystal structure of phenoxazine 4 bound to
both binding sites in TTR reveals that the simple design criteria
used to envision the N-phenyl phenoxazine inhibitors were by
in large correct (see results section). The structure confirms that
the molecular surface of 4 nicely complements the van der
Waals surface of the TTR binding cavity. The phenoxazine ring
system occupies the outer thyroid hormone binding cavity while
the m-CF3-substituted phenyl ring occupies the inner TTR
binding cavity. One of the two carboxylates comprising 4 is
close enough to make a favorable electrostatic interaction with
the ꢀ-ammonium side chain of Lys-15. As is typically the case
in structure-based ligand design, the structure of the complex
revealed some unexpected features as well. Phenoxazine 4 does
not enter the hormone binding channel as deeply as Flu,
suggesting that further modifications of the substructure project-
ing from N10 into the inner sub-site may improve binding and
efficacy, compare Supporting Information Figures 7 and 8. On
the other hand, deeper penetration of 4 could be deleterious, as
the high efficacy may be a direct result of interactions between
the phenoxazine ring system and the outer binding cavity of
TTR (vide infra). The other unexpected feature emerging from
the TTR‚(4)2 structure is that one of the carboxylates of 4 is in
close enough to share a proton with Glu-54; however the
energetic significance of this potential H-bonding interaction
for inhibitory activity is not yet known.
The molecular recognition event between TTR and 4 medi-
ated by complimentary van der Waals, hydrophobic, and
electrostatic interactions appears to prevent tetramer dissociation
and the linked conformational changes associated with TTR
amyloid formation (via ground-state stabilization and possible
transition state destabilization, see Figure 6).27,46,47 Due to the
increased size of the 4,6-phenoxazine dicarboxylic acid relative
to the anthranilic acid substructure of Flu, 4 is able to make
van der Waals contacts with both subunits of TTR composing
the binding cavity simultaneously. The isothermal titration
calorimetry data in Table 2 also suggest that inhibitor 4 and
Flu interact differently with TTR. The fact that the entropy of
(45) Andrea, T. A.; Cavalieri, R. R.; Goldfine, I. D.; Jorgensen, E. C.
Biochemistry 1980, 19, 55-63.
(46) Kelly, J. W.; Lansbury, P. T. J. Amyloid: Int. J. Exp. Clin. InVest.
1994, 1, 186-205.
(47) Blake, C. C. F.; Serpell, L. C.; Sunde, M.; Sandgren, O.; Lundgren,
E. In A Molecular Model of The Amyloid Fibril; Blake, C. C. F., Serpell,
L. C., Sunde, M., Sandgren, O., Lundgren, E., Eds.; John Wiley and Sons:
Chichester, 1996; Ciba Symposium Vol. 199, pp 6-21.

Design of TTR Amyloid Fibril Inhibitors
J. Am. Chem. Soc., Vol. 122, No. 10, 2000 2191
Figure 6. Free energy reaction coordinate diagram for the formation of the monomeric amyloidogenic intermediate from WT and L55P TTR (pH
4.4) in the presence (right side) and absence (left side) of 4. From previous studies we know that the WT tetramer dissociation is associated with
a high activation barrier (rate-determining step), which is decreased in the case of the L55P mutation.26 We also know from previous studies that
the WT tetramer is more stable that the L55P variant which is reflected in the free energy reaction coordinate diagram.8,9,11,25 Ligand binding is
known to decrease the free energy of the TTR‚(4)2 complex, making the activation barrier for L55P and WT TTR fibril formation more unfavorable.
If the transition-state structure mirrors the tetramer, as expected (Hammond Postulate), the T.S. energy will go up due to the binding of 4. Tetrameric
TTR is depicted in ribbon format with one of four identical subunits colored in gold bound to 2 molecules of 4 represented in space-filling format
(CPK). Note that the two exterior strands (the C and D strands shown in dark green) are presumed to melt away from the hydrophobic core upon
dissociation of tetrameric TTR yielding the six-stranded amyloidogenic intermediate which ultimately affords amyloid by self-assembly.
binding of 4 to TTR is favorable, while the entropy of Flu
binding is unfavorable can, in part, be attributed to the ability
of 4 to displace more H2O from the hydrophobic cavity, Table
2B. In addition, there is a minimal conformational entropy
penalty associated with the binding of 4 because it binds in its
lowest-energy conformation as discerned by a comparison of
the crystal structure of 22 and the bound conformation of 4 in
the TTR‚(4)2 structure, compare Figures 1D and 5A. The
conformational changes occurring in TTR (involving Ser-117
and Thr-119; vide infra) are very similar to those mediated by
Flu binding to TTR. Hence, the observed differences in the
enthalpy and entropy of ligand binding likely result because of
the different degree of penetration and molecular interactions
between 4 and TTR, and are not likely to be due to ligand-
induced conformational changes within TTR.
The non-cooperativity associated with the binding of 4 to
TTR relative to the negatively cooperative binding of Flu may
be a manifestation of the degree to which the ligands penetrate
the inner binding cavities. The deeper penetration of Flu is
envisioned to make solvent displacement more difficult upon
occupancy of the second binding site. Ligands such as T4 and
Flu which bind deep into the inner binding cavity exhibit
negative cooperativity, as opposed to compound 4 which binds
in the periphery of each cavity non-cooperatively. This is the
first structural clue that the degree of negative cooperativity
may be linked to the degree of difficulty associated with the
H2O displacement from the inner TTR cavity upon occupancy
of the second ligand binding site.
changes that may be important for preventing dissociation and
misfolding required for fibril formation (vide infra). Even though
the binding of 4 to TTR is not as energetically favorable as Flu
binding, the conformational changes mediated by 4 may be
different enough to make fibril formation more difficult. The
TTR conformational changes involving the Thr-119 and Ser-
117 side chains mediated by ligand binding not only allow for
increased van der Waals contacts between 4 and the protein,
but they also may stabilize the TTR tetramer and possibly
destabilize the transition state associated with misfolding.
Phenoxazine induced conformational changes in TTR also
facilitate the formation of a hydrogen bond between the Oγ
atom of Thr-119 to an ordered H2O molecule, which in turn is
hydrogen bonded to the carbonyl oxygen of Asp-18 on an
adjacent TTR subunit. The binding of 4 and Flu lead to a similar
desolvated hydrogen bonding network involving the Ser-117
residues on adjacent TTR subunits (see Introduction).33 Further
mutagenesis/biophysical studies are necessary to be confident
that these changes are energetically important and to address
whether they effect the ground state, the transition state, or both.
Conclusions
Starting with the 2.0 Å X-ray crystal structure of the
transthyretin‚(flufenamic acid)2 complex, a structure-based
ligand design strategy was utilized to conceive N-phenyl
phenoxazine transthyretin (TTR) amyloid fibril inhibitors.
Fifteen N-phenyl phenoxazines were chemically synthesized,
two of which proved to be excellent WT and L55P TTR amyloid
fibril inhibitors. The structure of one of the two most active
phenoxazines, 4, bound to TTR was solved to a resolution of
It is clear from both the WT-TTR‚(4)2 and WT-TTR‚(Flu)2
structures that these inhibitors mediate TTR conformational

2192 J. Am. Chem. Soc., Vol. 122, No. 10, 2000
Petrassi et al.
1.9 Å. N-phenyl phenoxazine 4 binds similar to the orientation
anticipated, although not as deeply into the channel as expected.
Ultracentrifugation studies demonstrate that 4 blocks the first
step of TTR amyloid fibril formation, that is, tetramer dissocia-
tion to the alternatively folded monomeric amyloidogenic
intermediate. It is clear that 4 functions in part by stabilizing
the normally folded tetramer through formation of the TTR‚
(4)2 complex, which in turn increases the activation energy for
tetramer dissociation. The data also suggest that 4 destabilizes
the transition state associated with TTR dissociation to the
monomeric amyloidogenic intermediate, Figure 6. The ability
to inhibit transthyretin amyloid fibril formation in vitro is
extremely important for enabling a test of the validity of the
amyloid hypothesis in vivo. The phenoxazine inhibitors 4 and
5 put us one step closer to that goal.
Welch Foundation (T.K., J.S.). We thank Mr. Scott Peterson,
Ms. Traci Walkup, Dr. Edward Koepf, and Ms. Linda Woo for
assistance with protein expression, Mr. Hans Purkey for carrying
out ITC studies on Flu binding to WT and L55P TTR and for
dissection of the energetics involved, and Mr. Hilal A. Lashuel,
Dr. Vibha Oza, and Dr. Prakash Raman for helpful discus-
sions involving analytical ultracentrifugation and chemical
synthesis.
Supporting Information Available: Included are the ITC
data and curve fits for 4 binding to WT and L55P TTR in
Figures 1 and 2, respectively; sedimentation velocity data on
WT and L55P TTR in the presence of 7.2 µM 4 at pH 4.4 in
Figures 3 and 4, respectively; ORTEP and stereoviews of 22 in
Figures 5 and 6, respectively; comparison of the degree of
penetration of Flu vs 4 into the TTR ligand binding site in
Figures 7 and 8, respectively; and Tables of crystallographic
data for 22 on several pages (PDF). This material is available
free of charge via the Internet at http://pubs.acs.org.
Acknowledgment. The authors gratefully acknowledge
primary financial support from the National Institute of Health
(R01 DK46335-01) and secondary support from The Skaggs
Institute of Chemical Biology, The Lita Annenberg Hazen
Foundation, the Donald and Delia Baxter Foundation, and the
JA993309V